FLEXIBLE TILE PARTITIONS

TECHNICAL FIELD

The present disclosure relates generally to video encoding and decoding techniques, and more particularly, to the encoding and decoding of flexible segments of a picture.

BACKGROUND

High Efficiency Video Coding (HEVC) is a block based video codec standardized by both the ITU-T and the Moving Picture Experts Group (MPEG) that utilizes both spatial and temporal prediction techniques. Spatial prediction reduces spatial redundancy and is achieved using intra (I) prediction from within the same frame of a current picture. Temporal prediction reduces temporal redundancy and is achieved using inter (P) or bi-directional inter (B) prediction on a block level using previously decoded reference pictures. Regardless of the particular type of prediction technique, however, the resultant differences between the original pixel data and the predicted pixel data, referred to as the“residual,” is transformed into the frequency domain and quantized. Quantizing the transformed residuals, the level of which is determined by the quantization parameter (QP), facilitates the control of a tradeoff between bitrate and the quality of the video.

The transformed and quantized residual is then entropy coded before being transmitted to a decoder together with the necessary prediction parameters. The prediction parameters, which are also entropy encoded, include prediction mode and motion vectors. Upon receipt, the decoder performs entropy decoding, inverse quantization, and inverse transformation to obtain the residual. The decoder then reconstructs the image from the residual using an intraprediction or inter-prediction technique.

Both MPEG and ITU-T are working on a successor to HEVC within the Joint Video Exploratory Team (JVET). The name of this video codec is Versatile Video Coding (VCC).

SUMMARY

Embodiments herein are useful for encoding and decoding a picture using flexible partitioning, for instance, partitioning that would reduce the costs of transmission of coded picture segments of the picture. For example, one or more embodiments enable a flexible partition structure where an edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments.

One or more embodiments include a method of decoding a picture. The method comprises receiving a bitstream. The bitstream comprises a plurality of coded picture segments and information associated with the plurality of coded picture segments. The method further comprises deriving, from the information in the bitstream, a partition structure. The partition structure divides the picture into picture segments. Each picture segment corresponds to one of the plurality of coded picture segments, and each picture segment comprises at least one unit. The edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The method further comprises decoding the plurality of coded picture segments received in the bitstream in accordance with the partition structure.

The derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure.

One or more embodiments include a corresponding method of encoding a picture that can be used in conjunction with or separately from methods for decoding a picture described herein. The method comprises defining a partition structure that divides a picture into picture segments. Each picture segment comprises at least one unit. An edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The method further comprises encoding the plurality of picture segments in accordance with the partition structure to generate a plurality of coded picture segments. Each coded picture segment corresponds to one of the picture segments of the partition structure. Each coded picture segment is independent such that derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure. The method comprises generating a bitstream comprising the plurality of coded picture segments and information indicating the partition structure used to divide the picture into the plurality of picture segments, and transmitting the bitstream.

Embodiments of the present disclosure also provide a decoder circuit configured to receive a bitstream. The bitstream comprises a plurality of coded picture segments and information associated with the plurality of coded picture segments. The decoder circuit is further configured to derive, from the information in the bitstream, a partition structure. The partition structure divides the picture into picture segments. Each picture segment corresponds to one of the plurality of coded picture segments, and each picture segment comprises at least one unit. The edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The decoder circuit is further configured to decode the plurality of coded picture segments received in the bitstream in accordance with the partition structure. The derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure.

Embodiments of the present disclosure also provide a decoder comprising receive circuitry configured to receive a bitstream comprising a plurality of coded picture segments and information associated with the plurality of coded picture segments, a memory comprising executable instructions, and processing circuitry operatively connected to the receive circuitry and the memory, and configured to execute the executable instructions to define a partition structure that divides a picture into picture segments. Each picture segment comprises at least one unit. An edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The processing is further configured to encode the plurality of picture segments in accordance with the partition structure to generate a plurality of coded picture segments. Each coded picture segment corresponds to one of the picture segments of the partition structure. Each coded picture segment is independent such that derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure.

The processing circuit is also configured to generate a bitstream comprising the plurality of coded picture segments and information indicating the partition structure used to divide the picture into the plurality of picture segments.

Embodiments of the present disclosure also provide an encoder circuit configured to define a partition structure that divides a picture into picture segments. Each picture segment comprises at least one unit. An edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The encoder circuit is further configured to encode the plurality of picture segments in accordance with the partition structure to generate a plurality of coded picture segments. Each coded picture segment corresponds to one of the picture segments of the partition structure. Each coded picture segment is independent such that derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure. The encoder circuit is further configured to generate a bitstream comprising the plurality of coded picture segments and information indicating the partition structure used to divide the picture into the plurality of picture segments, and transmit the bitstream.

Embodiments of the present disclosure also provide an encoder comprising a memory comprising executable instructions, processing circuitry operatively connected to the memory, and a transmitter. The processing circuitry is configured to define a partition structure that divides a picture into picture segments. Each picture segment comprises at least one unit. An edge of a first picture segment of the partition structure borders corresponding edges of two or more adjacent picture segments. The processing circuitry is further configured to encode the plurality of picture segments in accordance with the partition structure to generate a plurality of coded picture segments. Each coded picture segment corresponds to one of the picture segments of the partition structure. Each coded picture segment is independent such that derivation of any intra prediction mode for a first unit of the first picture segment depends on a derived intra prediction mode for a second unit of the first picture segment and is independent of any intra prediction mode for units of any other picture segments of the partition structure. The processing circuitry is further configured to generate a bitstream comprising the plurality of coded picture segments and information indicating the partition structure used to divide the picture into the plurality of picture segments. The transmitter is configured to transmit the bitstream

BRIEF DESCRIPTION OF THE DRAWINGS

Figure 1 illustrates an example of partitioning a CTU into CUs using QTBT.

Figure 2 illustrates an example tile partitioning.

Figures 3A-3B illustrates an example tiling structure with high resolution and low resolution tiles.

Figure 4 illustrates an example of a tile structure that is not supported in HEVC.

Figure 5 illustrates an example tile structure that can be supported in HEVC.

Figure 6 illustrates an example partitioning illustrating units and tiles.

Figure 7 illustrates an example bitstream according to one embodiment of the present disclosure.

Figure 8 illustrates a picture partitioned into 64 units.

Figure 9 illustrates an example three segment partition structure according to one embodiment of the present disclosure.

Figure 10 illustrates a picture that has been partitioned using a partition structure illustrated in Figures 8 and 9 according to one embodiment of the present disclosure.

Figures 1 1A-1 1 B illustrate a picture partitioned according to a partition structure, with Figure 1 1A illustrating the units, and Figure 1 1 B illustrating the segments or tiles, according to one embodiment of the present disclosure.

Figure 12A is a flow diagram illustrating a method for encoding a picture or part of a picture using flexible tiles in accordance with a first embodiment of the present disclosure.

Figure 12B is a flow diagram illustrating a method for decoding a picture or part of a picture that was encoded using flexible tiles in accordance with the first embodiment of the present disclosure.

Figure 13A is a flow diagram illustrating a method by which the encoder constructs and signals the flexible tiles according to one embodiment of the present disclosure.

Figure 13B is a flow diagram illustrating a method by which the decoder parses and constructs the flexible tiles according to one embodiment of the present disclosure.

Figure 14 illustrates an example tile partition where 10 tiles are used to partition the picture according to one embodiment of the present disclosure.

Figure 15 illustrates an example tile partition in which a picture having a size of 2048x1280 is partitioned into 26 tiles according to one embodiment of the present disclosure. Figure 16A illustrates a method implemented at a decoder for decoding an encoded representation of a picture of a video bitstream according to one embodiment of the present disclosure.

Figure 16B illustrates a method implemented at an encoder for encoding a

representation of the picture of a video bitstream according to one embodiment of the present disclosure.

Figure 17A-17C illustrate a tile partition of the picture illustrated in Figure 14 and corresponding ctu_status bits according to one embodiment of the present disclosure.

Figure 18A is a flow diagram illustrating a method implemented at an encoder for constructing and signaling the tiles according to one embodiment of the present disclosure.

Figures 18B-18C illustrates a method implemented at a decoder for parsing and constructing the tiles sent by the encoder according to one embodiment of the present disclosure.

Figures 19A-19B are flow diagrams illustrating methods implemented by an encoder and a decoder, respectively, to construct and signal tiles according to one embodiment of the present disclosure.

Figure 20 illustrates the integer division of tiles according to one embodiment of the present disclosure.

Figures 21A-21 B illustrate a method implemented by an encoder to construct and signal tiles according to one embodiment of the present disclosure.

Figures 21 C-21 D illustrate a method implemented by a decoder to parse the tiles constructed and signaled by the encoder in Figures 21A-21 B according to one embodiment of the present disclosure.

Figures 22A-22C illustrate horizontal and vertical divider lines corresponding to the tile structure of Figure 14 according to one embodiment of the present disclosure.

Figure 23A is a flow diagram illustrating a method implemented at the encoder for determining the horizontal and vertical divider lines for a given tile structure according to one embodiment of the present disclosure.

Figure 23B is a flow diagram illustrating a method implemented at the decoder for processing tiles having a given tile structure described with corresponding horizontal and vertical divider lines according to one embodiment of the present disclosure.

Figures 24A-24C illustrate an example tile structure and the corresponding set of T- junctions before and after removing redundant T-junctions according to one embodiment of the present disclosure.

Figures 25A-25B illustrate an example tiling structure and a corresponding set of T- junctions according to one embodiment of the present disclosure.

Figure 26A is a flow diagram illustrating a method implemented at the encoder for creating the set of T-Junctions according to one embodiment of the present disclosure. Figure 26B is a flow diagram illustrating a method implemented at the decoder for creating the set of T-Junctions according to one embodiment of the present disclosure.

Figures 27A-27C illustrate an exemplary tiling structure and the corresponding matrix of start corners and matrix of end corners according to one embodiment of the present disclosure.

Figure 28A is a flow diagram illustrating a method implemented at the encoder for creating the start and end corner matrices according to one embodiment of the present disclosure.

Figure 28B is a flow diagram illustrating a method implemented at the decoder for processing the start and end corner matrices according to one embodiment of the present disclosure.

Figure 29 illustrates a system configured to exchange a series of images that are encoded and decoded according to one embodiment of the present disclosure.

Figure 30 is a schematic block diagram illustrating components of an encoder.

Figure 31 is a schematic block diagram illustrating components of a decoder

Figures 32-33 are schematic block diagrams illustrating some functional components of an encoder and a decoder, respectively.

Figure 34A is an encoder encoding an image according to a partition structure according to one embodiment of the present disclosure.

Figure 34B is a decoder decoding an image according to a partition structure according to one embodiment of the present disclosure.

Figure 35 is a system comprising a decoder and encoder according to one embodiment of the present disclosure.

DETAILED DESCRIPTION

This application claims priority of U.S. Application No. 62/688314, filed 21 June 2018, disclosure of which is incorporated in its entirety by reference herein.

Quad Tree and Binary Tree (QTBT) structure

As previously stated, HEVC is a block based video codec standardized by ITU-T and MPEG that utilizes both temporal and spatial prediction. HEVC uses a block structure where each top level coding block, i.e. the largest block in the coding block partitioning referred to herein as the Coding Tree Unit (CTU), is partitioned by a Quad Tree (QT) structure. This partitioning produces coding block partitions, referred to herein as coding units (CUs). The CUs can then be further partitioned, recursively, into smaller equally sized CUs with the quad tree structure down to an 8x8 block size.

The block structure in the current version of VVC is different than the block structure in HEVC. Particularly, the block structure in VVC is referred to as Quad Tree plus Binary Tree plus Ternary Tree block structure (QTBT+TT). A CU in QTBT+TT can have either square or rectangular shapes. A coding tree unit (CTU) is first partitioned by a quad tree structure as in HEVC. Then the CTU is further partitioned with equally sized partitions either vertically or horizontally in a binary structure to form coding blocks (also referred to as CUs). A block could thus have either a square or rectangular shape.

The depth of the quad tree and the binary tree can be set by the encoder in the bitstream. An example of dividing a CTU using QTBT+TT is illustrated in Figure 1. The TT allows the CU to be divided into three partitions instead of two equally sized partitions. This increases the possibility of using a block structure that better fits the content structure in a picture.

Context Adaptive Binary Arithmetic Coding (CAB AC)

Context Adaptive Binary Arithmetic Coding (CABAC) is an entropy coding tool used in HEVC and VVC. CABAC is configured to encode binary symbols, which keeps complexity low and allows modelling of probabilities for bits of a symbol that are more frequently used. The probability models are selected adaptively based on local context, since coding modes are usually well correlated locally.

Slices

The concept of slices in HEVC divides the picture into independently coded slices, where each slice is read in raster scan order in units of CTUs. Different coding types could be used for slices of the same picture. For example, a slice could either be an l-slice, P-slice or B- slice. However, the main purpose of slices is to enable resynchronization in case of a loss of data.

Tiles

The HEVC video coding standard also includes a tool called“Tiles” that divides a picture into rectangular, spatially independent regions. Using tiles, a picture in HEVC can be partitioned into rows and columns of samples where any given tile is located at an intersection of a given row and a given column. Figure 2 shows an example of a tile partitioning using four (4) rows of tiles and five (5) columns of tiles resulting in a total of twenty (20) tiles for the picture. As seen in Figure 2, the tiles in HEVC are always aligned with CTU boundaries.

The tile structure is signaled in the Picture Parameter Set (PPS) by specifying the thicknesses of the rows and the widths of the columns. Individual rows and columns can have different sizes, but the partitioning always span across the entire picture, from left to right and top to bottom respectively.

The PPS syntax used for specifying the tile structure in HEVC is listed below in Table 1. As seen in Table 1 , a flag (i.e. , the tiles_enabled_flag) indicates whether tiles are used or not used. If the tiles_enabled_flag is set, the number of tiles columns (i.e.,

num_tile_columns_minus1) and rows (i.e., num_tile_rows_minus1) are specified. The uniform_spacing_flag is a flag that specifies whether the column widths and row heights are explicitly signaled, or whether a pre-defined method to space the tile borders evenly should be used. If explicit signaling is indicated, the column widths are signaled one-by-one followed by the row heights. The column width and row height information is signaled in CTU units. Finally, the flag loop_filter_across_tiles_enabled_flag specifies whether in-loop filters across tile boundaries are turned on or off for all tile boundaries in the picture.

pic_parameter_set_rbsp( ) { Descriptor tiles_enabled_flag u(1 ) if( tiles_enabled_flag ) {

num_tile_columns_minus1 ue(v) num_tile_rows_minus1 ue(v) uniform_spacing_flag u(1 ) if( !uniform_spacing_flag ) {

for( i = 0; i < num_tile_columns_minus1 ; i++ )

column_width_minus1 [ i ] ue(v) for( i = 0; i < num_tile_rows_minus1 ; i++ )

row_height_minus1 [ i ] ue(v)

}

loop_filter_across_tiles_enabled_flag u(1 )

}

Table 1 - Tile syntax in HEVC

Similar to slices, there is no decoding dependency between tiles of the same picture.

This includes intra prediction, context selection and motion vector prediction. One exception, however, is that in-loop filtering dependencies are generally allowed between tiles. Those dependencies can, however, be disabled by setting the loop_filter_across_tiles_enabled_flag appropriately.

In contrast to slices, tiles do not require as much header data. The header overhead per tile comprises the signaling of bit-stream offsets, which are present in the slice header and indicate the starting point of all tiles in a picture. A decoder decodes the starting points to enable splitting the coded picture into coded tiles in order to distribute them for parallel decoding. In HEVC, the inclusion of bit-stream offsets in the slice header is mandatory when tiles are enabled. However, the combination of tiles and slices is restricted in HEVC. Particularly, either all CTUs in a tile belong to the same slice or all CTUs belong to the same tile.

Bit-stream offsets can also allow the extraction of tiles and stitching of tiles to re constitute the tiles into an output stream. This requires some encoder side constraints to make the tiles temporally independent. One constraint restricts motion vectors so that motion compensation for a tile only uses samples included in spatially co-located tiles of previous pictures. Another constraint restricts temporal motion vector prediction (TMVP) such that this process is made temporally independent. For complete independence, the deblocking of boundaries between tiles must be disabled via the previously described

loop_filter_across_tiles_enabled_flag. However, disabling deblocking may introduce visible lines between tiles. Therefore, some implementations disable deblocking while other

implementations do not.

Tiles are sometimes used for 360-degree video that is intended for consumption using head-mounted display (HMD) devices. The field-of-view when using today’s HMD devices is limited to around 20% of a full sphere. This means that only 20% of the full 360-degree video is consumed by the user. Typically, the entire 360-degree video sphere is made available to the HMD device, which then crops out the part that is rendered for the user. That part, i.e. the part of the sphere the user sees, is called the viewport. A well-known optimization of resources is to make the HMD device video system aware of head movements and the direction the user is looking so that fewer resources are spent on processing video samples that are not rendered to the user. The resources can be, for example, bandwidth from a server to a client or the decoding capability of the device. For future HMD devices, where the field of view will be larger than is currently possible, a non-uniform resource allocation would still be beneficial.

Particularly, the human vision system demands a higher image quality in the central vision area (about 18 ^° horizontal view), while a lower demand is placed on the image quality in the peripheral region (about 120 ^° or more for a comfortable horizontal view). Therefore, non- uniform resource allocation would be helpful to meet the demands of the human vision system with more resources being allocated in the central vision area as compared to the peripheral region.

Optimizing resources to the Region of Interest (Rol) is another use case for tiles. Rol can be specified in the content or extracted by methods such as eye tracking. One method of using head movements to reduce the amount of required resources is to use tiles. This method first encodes the video sequence multiple times using tiles. The tile partitioning structure is the same in all encodings; however, the video sequence is encoded at different video qualities.

This produces at least one high-quality encoding for the video sequence and one low-quality encoding for the video sequence. This means that for each tile at a particular point in time, there are at least one high-quality tile representation and at least one low-quality tile representation. The difference between a high-quality tile and a low-quality tile can be that the high-quality tile is encoded at a higher bitrate than the low-quality tile, or that the high-quality tile is of higher resolution than the low-quality tile.

Figures 3A-3B illustrate an example of high-quality tiles having higher resolution than low-quality tiles. Particularly, Figure 3A illustrates a picture that just has been decoded by a decoder. In this context, pictures showing the output from the decoder (i.e., the decoded pictures) are pictures shown in the compressed domain. In this example, all 8 tiles A-H in the compressed domain are of equal spatial size. Then, after the picture is decoded, but before the picture is rendered, the tiles A-H are scaled and arranged spatially. Pictures that have been prepared for rendering (i.e., after scaling and re-arrangement) are pictures to be shown in the output domain. The output domain, seen in Figure 3B, illustrates the picture as it is rendered or presented to the user.

As seen in Figures 3A-3B, tiles B, D, F, and H are high-quality tiles since they are of higher resolution in the output domain. Tiles A, C, E, and G, however, are low-resolution tiles since the scaling step decreases the actual resolution.

Besides illustrating how tiles have different resolutions, Figures 3A-3B also illustrate that tiles in the compressed domain (Figure 3A) do not need to be spatially ordered in the same way as they are ordered in the output domain (Figure 3B). Given that the tiles are both spatially and temporally independent, the spatial placement of tiles in the compressed domain is not vital.

There are multiple elements that increase bit-costs when enabling tiles in HEVC. First, prediction across tiles is disabled, which means that motion vectors and intra modes are not predicted across tiles. The use of tiles also disables quantization parameter (QP) prediction and context selection. Second, CABAC is initialized for each tile, which means that CABAC adaptation is impaired. Third, bit-stream offsets must be signaled for each tile. Fourth, the tile partitioning structure needs to be specified in the PPS. Finally, CABAC is flushed after each tile and the coded data has to be byte-aligned.

Tiles are useful; however, there are some considerations that need to be addressed.

For example, in its current form, HEVC restricts tiles to ensure that they span across the entire picture. This limits the flexibility of tiles, however. For example, Figure 4 illustrates an example of a tile structure that is not supported by the current implementation of HEVC. In order to virtually support such a tile structure, one may specify the partition structure in HEVC seen in Figure 5. However, the tile structure seen in Figure 5 comprises more tile boundaries. This can harm compression efficiency and cause unnecessary tiling artifacts. Additionally, if the tiles are to be completely spatially independent, HEVC in its current form only allows for either enabling or disabling the in-loop filtering dependencies for all tile boundaries of a picture. Therefore, unnecessary in-loop filtering restrictions will occur with conventional HEVC solutions because they do not support flexible tile partitions.

Embodiments of the present disclosure address these issues by providing flexible tile partitioning methods. These methods provide more freedom for an encoder to partition pictures into tiles. In one embodiment, each tile comprises a single rectangular area, and the width and height of each tile is signaled to a decoder in a bitstream. Upon receipt, the decoder is configured to decode the individual width and height values for each tile from the bitstream. In another embodiment, the picture is divided into units, with each unit being assigned a tile ID that identifies which particular tile the unit is assigned to. This latter embodiment is beneficial as it supports flexible tile partitions having non-rectangular tiles, as well as partitions in a

checkerboard pattern. Embodiments of the present disclosure enable the use of more flexible tile partitions for various use cases, such as 360-degree video. Further, when compared with fixed tile partitions, the bit cost associated with flexible tile partitions is lower. Specifically, the number of tiles is higher with the use of conventional tile partitioning, and each tile comes at a cost (e.g. the processing associated with CABAC flush and adaptation reset, bit-stream offsets, etc.). The flexible tile partitioning of the present embodiments, however, uses fewer tiles thereby reducing bit costs.

The present embodiments also enable flexible CPU load balancing for parallel processing where an encoder can, for example, analyze an input picture and select a tile partitioning that splits the picture into as few tiles as possible (to reduce bit cost) while filling the available number of cores equally. Moreover, embodiments of the present disclosure also facilitate increased control over in-loop filters as compared to conventional methods that to make traditional tiles more flexible.

The present embodiments are even more beneficial for demanding use cases, such as Multi-View (MV) video coding, for example, where the bandwidth or decoding capability is a major bottleneck. In such cases, the present embodiments enable better resource allocation. In applications where the region of interest (Rol) is specified, the embodiments of the present disclosure also provide a more flexible resource allocation.

The flexible tile partitioning embodiments of the present disclosure use larger tiles than do conventional tiling methods. Because of the larger tile size, the present embodiments also contribute to the quality of a picture by putting fewer limitations on the intra motion vectors. Further the flexible tile partitioning methods of the present disclosure removes unnecessary partitioning lines between tiles. Such line removal reduces the cost of filtering with respect to lines between the tiles.

Embodiments of the present disclosure partition a picture from a sequence of video pictures into tiles with a more flexible tile layout than can be provided by conventional means, e.g. HEVC. This includes configuring an encoder to generate the partition structure and tiles for encoding, as well as a decoder to decode the partition structure and tiles. In some

embodiments, the flexible tile partitioning method of the present disclosure co-exists with the processing of HEVC in its conventional form. Thus, as previously described, the

tiles_enabled_flag in the bitstream can be used to specify whether the traditional tile partitioning scheme should be used, or whether the flexible tile partitioning scheme according to the present embodiments should be used.

It should be noted that some embodiments refer to the term“units.” As defined herein, a unit is rectangular area of the picture (e.g. coding tree units or coding units) such that a tile consists of one or multiple units. Figure 6 illustrates an example in which a picture 10 is partitioned into four (4) tiles or“segments” T-i, T ₂ , T3, T4 (collectively,“T”) (as shown by the solid black lines), with each tile T comprising 16 units U. This provides a total of 64 units U for the picture 10.

The flexible tile structure is signaled in a bitstream 12. For example, the encoder signals the flexible tile structure in a parameter set such as the sequence parameter set (SPS) or picture parameter set (PPS). However, as those of ordinary skill in the art will appreciate, the present disclosure is not limited to signaling the flexible tile structure in a parameter set.

Although the embodiments herein describe the tile structure as being signaled in the PPS, this is for illustrative purposes only and this information can be signaled in other ways.

General Description

Figure 7 shows a bitstream 12 that is decoded into a picture 10. Figure 8 shows the picture 10 partitioned into 64 units U. One example of a unit U is a macroblock known from H.264. Another example of a unit U is a CTU known from HEVC.

As seen in Figure 7, bitstream 12 comprises one or more coded segments 14 and one or more code words 16, 18. The one or more codewords 16 comprises information that specifies the number of bytes or bits in the bitstream comprising the data needed for decoding each of the coded segments 14. The one or more codewords 18 carry information specifying a spatial size of the units U. As an example, the spatial size of the units is 64 luma samples horizontally and 64 luma samples vertically; however, as those of ordinary skill in the art will appreciate, the present disclosure is not limited to these sizes and shapes of units. Other unit sizes and shapes, including irregularly-shaped units (e.g., units not shaped as squares), are also possible according to the embodiments of the present disclosure.

As seen in Figures 7-8, picture 10 has been spatially partitioned into three segments T-i , T2, T3 (collectively,“T”) using a partition structure S. The information defining the partition structure S was derived from information in the bitstream 12. For example, the coded segments 14 from the bitstream 12 are decoded to derive the sample values of the corresponding decoded segment in the picture 10. Additionally, the size and/or position of the coded segment 14 can be parsed from the one or more codewords 16 in the bitstream 12. As previously stated, one embodiment of the present disclosure utilizes the one or more codewords 16 to carry information specifying the number of bytes or bits in the bitstream that comprise information needed to decode the coded segment 14. Alternatively, however, embodiments of the present disclosure can utilize the one or more codewords 16 to carry information needed to derive the locations or positions of the coded segment 14 in the bitstream 12.

As seen in Figure 8, the boundaries of each segment T are aligned with the boundaries of the unit U. This means that each segment T comprises an integer number of units U, and that each unit U is associated with one segment T. For example, as seen in Figure 8, the two rightmost segments T 1 , T2 comprise 16 units each, while the leftmost segment T3 comprises 32 units U. Figure 9 illustrates an embodiment in which picture 10 is partitioned using the same partition structure S illustrated in Figure 8. However, for clarity’s sake, Figure 9 does not illustrate the units U. As seen in Figure 9, one side 19-3 of a segment T3 borders the sides 19- 1 , 19-2 of two or more neighboring or adjacent segments T 1 , T2, respectively.

Figure 10 illustrates a picture 10 that has been partitioned using the same partition structure S illustrated in Figures 8 and 9. As seen in Figure 10, picture 10 comprises three segments T ₁ , T ₂ , T ₃ . All segments T ₁ , T ₂ , T ₃ of Figure 10 comprise one or more units U;

however, for clarity sake, Figure 10 illustrates segments T ₁ and T ₃ as comprising a subset of the units U. Specifically, segment T ₃ comprises unit Ui while the neighboring or adjacent segment Ti comprises units Ui, U2. Each of the segments T ₁ , T ₂ , T ₃ are independent with respect to each other. This means that the boundaries of the segments T ₁ , T ₂ , T ₃ are handled similar to picture boundaries when decoding the units U. This affects the derivation process of elements during decoding such as, e.g., the derivation of intra prediction modes and the derivation of quantization parameter values.

Intra prediction modes (e.g., intra (B) prediction mode and intra (P) prediction mode) are used and signaled for units that only use predictions from previously decoded samples of a current picture for sample prediction. It is common that the derivation of the intra prediction mode in a particular unit Ui depends on previously derived intra prediction modes in other units U ₂ . With segments being independent, the derivation of the intra prediction mode in a particular unit Ui only depends on previously derived intra prediction modes in units U ₂ that belong to the same current segment T1 and does not depend on any intra prediction mode in any unit, such as unit U ₃ , that is associated with a different segment, such as segment T ₃ . Thus, the partition structure S of Figure 10 makes the intra prediction modes in the units U ₃ in segment T ₃ unavailable for the derivation of the intra prediction mode for units Ui , U ₂ in a different segment T 1 .

Unit U ₃ in segment T ₃ could, however, be used for derivation of an intra prediction mode in another unit, but only if that other unit was in the same segment T ₃ . Thus, the segment boundaries have the same or similar effect on intra mode derivation as a picture boundary for the units Ui, U2 in second segment Ti.

The quantization parameter (QP) value is also well known in the art. The QP value for a unit U, for example, is typically derived using previously derived QP values. One derivation process known in the art is to first derive a predicted QP value from the bitstream 12, and then add a delta QP value to the predicted QP value. Because the segments T-i , T ₂ , T ₃ are independent of each other, the derivation of the QP value in a particular unit, such as unit U ₂ of segment T 1 , for example, can only depend on previously derived QP values in units that are in the same segment T ₁ . The derivation of a QP value in a particular given unit Ui , U ₂ does not depend on any QP values associated with any other unit that is in a different segment, such as unit U ₃ in segment T ₃ . This means that the partition structure S in Figure 10 makes the QP values in units U ₃ in segment T ₃ unavailable for the derivation of the QP value for the units Ui,

U ₂ in a different segment T ₃ .

The QP value in some units U could be used for the derivation of a QP value in another unit U, but only if that other unit was in the same segment Ti. Thus, the segment boundaries have the same or similar effect on QP value derivation as a picture boundary for the units Ui , U2 in segment Ti .

Figures 1 1A-1 1 B illustrate the same picture 10 partitioned according to the partition structure S. Figure 1 1 A illustrates the units U. As previously described, the spatial size of the units is signaled in the bitstream 12 as one or more code words 16, 18. Figure 1 1 B illustrates segments T ₁ , T ₂ , T ₃ and the partition structure S of picture 10. Each segment T ₁ , T ₂ , T ₃ comprises at least one unit U. The spatial size of each segment T ₁ , T ₂ , T ₃ is expressed in terms of units, which means that the size of the units U is used when deriving the spatial sizes of the segments and/or the spatial partition structure. In the example of Figures 1 1 A-1 1 B, the spatial size of two of the segments T 1 and T2 is 4 units horizontally and 4 units vertically. The spatial size of one segment is T ₃ units horizontally and 8 units vertically.

It should be noted that segments T ₁ , T ₂ , T ₃ in some embodiments can be rectangular such that the width and height of the segments differ. In other embodiments, segments T ₁ , T ₂ , T ₃ are non-rectangular and expressed by allocating each unit U to one segment. Such an allocation is used to represent, for example,“chessboard segments,” in which some of the segments T represent white squares on a chessboard, while other segments T represent the black squares on the chessboard. A given segment T can, in some embodiments, be equivalent to a tile or slice.

First Embodiment

As seen in a first embodiment in Figure 9, a picture 10 or a part of a picture 10 in a video stream is divided into a spatial partition structure S, where the partition structure S comprises at least three segments T ₁ , T2, T ₃ . Each segment comprises units U, with at least one segment T, such as segment T 1 , for example, comprising two or more units U. The spatial size of the units U is signaled in, and decoded from, the one or more codewords 18 of the bitstream 12.

Additionally, the one or more code words 16 indicating the size or position of at least one coded segment 18 in the bitstream 12 are parsed from the bitstream 12.

At least one segment T ₃ has at least one side 19-3 that borders more than one neighboring segment T ₁ , T ₂ . Additionally, the segments T ₁ , T ₂ , T ₃ are independent with respect to each other such that the derivation of any intra prediction mode for any unit U, such as unit Ui in a segment Ti depends only on previously derived intra prediction modes in units U ₂ that is also in the segment Ti. The derivation of any intra prediction mode for a unit Ui in the segment Ti does not depend on any intra prediction mode in any unit U ₃ that is in a different segment T ₂ . Such segments are referred to herein as“flexible tiles” or simply just“tiles,” and are different from HEVC tiles which are arranged in as strict columns and rows. That is, with HEVC, no HEVC tiles in the picture have one side that borders more than one neighboring HEVC tile. Rather, a given side of a given HEVC tile borders only one side of one other adjacent HEVC tile.

It should be noted that the present disclosure uses the terms segments, flexible tiles, and tiles interchangeably. However, within the context of these embodiments, the terms “flexible tile,” and“tile” are defined to be the same as a segment, and all are different than HEVC tiles, as previously described.

Figure 12A is a flow diagram illustrating a method 20 for encoding a picture or part of a picture using flexible tiles in accordance with a first embodiment of the present disclosure.

Method 20 is implemented at an encoder and begins with the encoder defining a flexible tile structure for the picture (box 22). The flexible tile structure comprises at least one flexible tile or“segment” having at least one side that borders two or more adjacent flexible tiles or “segments.” The encoder then defines at least one syntax element to describe how to construct the flexible tile structure (box 24). The at least one syntax element comprises one or more properties that describe the flexible tiles, or the syntax element comprises means to derive properties for the flexible tiles. Such properties include, for example, an identifier, a width, a height, and a position that describe how to assign and/or access the individual flexible tiles. The encoder then signals the at least one syntax element for the flexible tile structure in a video bitstream (box 26), encodes the picture or part of the picture according to the flexible tile structure into the video bitstream (box 28), and transmits the bitstream 12 (box 29).

Figure 12B is a flow diagram illustrating a method 30 for decoding a picture or part of a picture that was encoded using flexible tiles in accordance with the first embodiment of the present disclosure.

As seen in Figure 12B, method 30 begins with the decoder receiving a video bitstream (box 32) and parsing the at least one syntax element for the flexible tile structure from the video bitstream received from the encoder (box 34). The decoder can then construct the flexible tile structure from the at least one syntax element, and assign or derive properties for the individual flexible tiles (box 36). The properties include, for example, an identifier, a width, a height, and a position that is used for assigning and/or accessing the flexible tiles. The decoder then decodes the picture or part of the picture according to the flexible tile structure (box 38). As stated previously, the flexible tile structure comprises at least one flexible tile having at least one side that borders corresponding sides of more than one adjacent flexible tile.

Second Embodiment

In a second embodiment, the flexible tiles are scanned in a pre-defined order, and a tile width and tile height are specified for each tile. For instance, the pre-defined order is a raster scan order, e.g. along the CTUs, and the tile size is specified in CTU units.

Figure 13A is a flow diagram illustrating a method 40 by which the encoder constructs and signals the flexible tiles. As seen in method 40, the encoder first determines whether the picture has one tile or multiple tiles (box 42). If the picture has only one tile, the encoder sets the tile_enabled_flag equal to 0 to indicate that the picture has only a single tile (box 44) and ends the process. Otherwise, if the picture has more than one flexible tile, the encoder creates an empty list of tiles (box 46). Following a predefined scan order, the encoder adds the flexible tile to the list, including the information identifying the height and width of the flexible tile (box 48). The encoder continues to add flexible tiles to the list as long as the picture has more flexible tiles (box 50).

Figure 13B is a flow diagram illustrating a method 60 by which the decoder parses and constructs the flexible tiles.

As seen in method 60, the decoder first determines whether the picture has one tile or multiple tiles (box 62). If the tile_enabled_flag is set to 0, indicating that the picture has only one flexible tile with the identified width and height, end the process. Otherwise, while there are still flexible tiles in the list (box 64), parse the next tile in the list, specified by its width and length following the predefined scan order (box 66).

Table 2 identifies an example syntax for use in one or more embodiments of the present disclosure. The syntax could be implemented on top of the currently existing HEVC video coding specification.

able 2 - An example tile syntax

Semantics for the Syntax Elements in Table 2

tiles_enabled_flag: This flag can be set to a value of 1 or 0. When set to 1 , this flag specifies that there is more than one tile in each picture referring to the PPS. When set to 0, this flag specifies that there is only one tile in each picture referring to the PPS. tile_width_minus1 [ i ] plus 1 : specifies the width of the i-th tile in units of coding tree units. tile_height_minus1 [ i ] plus 1 : specifies the height of the i-th tile in units of coding tree units. read_more_tiles( ):

set the return value of read_more_tiles( ) to TRUE if there is at least one CTU in the picture that is not assigned to a tile.

Otherwise, set the return value of read_more_tiles( ) to FALSE.

In one embodiment, the decoder invokes the following function prior to parsing the first tile_width_minus1 syntax element in a PPS:

tilejd = 0

i = 0

for( y= 0; y < PicHeightlnCtbsY; y++ )

for( x = 0; x < PicWidthlnCtbsY; x++ )

CTUJile_assignment[ y ][ x ] =“no tile”

In one embodiment, the decoder invokes the following function after parsing the first tile_width_minus1 syntax element in a PPS.

for( tile_coordinate_y = 0; tile_coordinate_y < PicHeightlnCtbsY; tile_coordinate_y++ )

for( tile_coordinate_x = 0; tile_coordinate_x < PicWidthlnCtbsY; tile_coordinate_x++ ) if( CTUJile_assignment[ tile_coordinate_y ][ tile_coordinate_x ] ==“no tile” ) break

for( y = 0; y <= tile_height_minus1 ; y++ )

for( x = 0; x <= tile_width_minus1 ; x++ )

CTUJile_assignment[ tile_coordinate_y+y][tile_coordinate_x+x] = tilejd tile id++

i++

In one embodiment, the syntax elements tile_width_minus1 and tile_height_minus1 are restricted such that all of the following conditions must hold prior to CTU_tile_assignment

[ tile_coordinate_y+y][tile_coordinate_x+x] = tilejd:

1. the tile_coordinate_y+y must be less than PicHeightlnCtbsY;

2. tile_coordinate_x+x must be less than PicWidthlnCtbsY; and

3. CTU_tile_assignment[ tile_coordinate_y+y][tile_coordinate_x+x] must be equal to“no tile.”

Figure 14 illustrates an example tile partition where 10 tiles To...T9 are used to partition the picture 10. The numbers seen in each tile of Figure 14 represent a tile number (i.e., an integer 0, 1 , 2,...) followed by the width and height of the tile in CTUs (e.g., 4x2). Using the example syntax of Table 2 above, the tile partition is expressed by a total of 63.

Figure 15 shows another example tile partition a picture 10 having a size of 2048x1280 is partitioned into 26 tiles. As seen in Figure 15, not all tiles are the same size and shape. Rather, there is one large sized tile TL and a plurality of smaller tiles Ts. The large rectangular tile TL in Figure 15 is a flexible tile as it has one side that borders the sides of more than one of the smaller tiles Ts. Additionally, the picture comprises 32x20 = 640 CTUs, each of which has a size of 64x64 luma samples. Such a tile partition is used for 360 video streaming where some smaller tiles T _s are used to cover the viewport and a single large tile TL covers the full sphere in lower resolution.

Using the example syntax in Table 2 above, the tile partition of Figure 15 can be expressed by a total of 267 bits. One bit for the tiles_enabled_flag =1 indicating that flexible tiles are used. Then the 25 smaller tiles Ts are signaled with 10 bits each, and the large tile TL is signaled with 16 bits using tile_width_minus1 = 1 1 (7 bits) and tile_height_minus1 = 19 (9 bits).

If the tile size is used, the result would be 1 + tile size (5+5) + 25 ^* 2 + (3+5) = 69 bits.

Third Embodiment

In a preferred embodiment, the tile partition structure is conveyed by signaling the height and width of tiles. The tiles consist of an integer number of coding units such as CTUs. The spatial size of the coding units is signaled in the bitstream.

For example, a CTU unit size is signaled in the bitstream to have a size of 64x64 luma samples. One tile is then signaled to have a width of 2 and a height of 1 , which means that the tile has a width of 128 luma samples (i.e., 2x64 luma samples) and a height of 64 luma samples (i.e. , 1 x64 luma samples).

Further, in this embodiment a flag is used to specify whether or not a current tile size is identical to the closest previously signaled tile size. If the flag has one value, the tile size for the current tile is copied from a previous tile. If the flag has another value, the height and width of the tile is explicitly signaled using the height and width of the coding unit. As described in more detail below, the decoder decodes this flag and, based on its value, determines whether or not the current tile size is identical to the closest previously signaled tile size.

In one variant of this embodiment, the number of tiles in the picture 10 is decoded first. Then there is a loop over the known number of tiles for which the size is decoded. Optionally, the size of the last tile is not signaled since the size of that last tile is given by the remaining area of the picture.

In another variant of this embodiment, the number of tiles in the picture is not signaled. Instead, the decoder continues decoding tile sizes until the decoder determines that all the necessary tile sizes have been decoded. When all necessary tile sizes have been decoded, the decoder stops decoding tile sizes and proceeds decoding other data from the bitstream. One way to determine whether all necessary tile sizes have been decoded is to determine whether all CTUs in the picture are in a tile for which the tile size has been decoded. Another way to determine whether all necessary tile sizes have been decoded is to determine whether the tiles corresponding to the decoded tile sizes jointly cover the entire picture.

In yet another variant of this embodiment, an optional codeword is signaled to convey the number of times a tile is being copied to the next tile position. For example, a first tile size is equal to 128x128. Then there can be one or more codewords in the bitstream that specifies whether that tile size:

1 ) is not copied to any other tile, which means that the next tile size is signaled explicitly using a width and height; or

2) is copied to the next tile; or

3) is copied to the next n tiles, where the number of times to copy is signaled.

As seen in Figure 16A, a decoder is configured to decode an encoded representation of a picture of a video bitstream using one or more of the following aspects.

Particularly, as seen in method 70, the decoder receives coded elements with which to derive a specific spatial unit size for decoding the picture. Alternatively, the decoder is configured to use a specific spatial unit size to decode the picture (box 72). The decoder can also decode more than one tile size explicitly from the video bitstream (box 74). In these cases, the tile size comprises a width value that is derived from the bitstream as an integer number of the unit width, as well as a height value that is derived from the bitstream as an integer number of the unit height. In one embodiment, the decoder can decode at least one tile size from the video bitstream, wherein the tile size comprises a width value and a height value that is copied from a previously decoded tile size (box 76). In one embodiment, the decoder is configured to decode a flag for at least one current tile (box 78). In these cases, one value could indicate that the tile size of the current tile is copied from a previously decoded tile size, while another value indicates that the tile size should be explicitly decoded from information in the bitstream. In one embodiment, the decoder determines the number of tile sizes to decode by decoding a number value from the bitstream specifying the number of tile sizes (box 80). In one embodiment, the decoder determines the number of tile sizes to decode by determining whether there are additional tile sizes to decode (box 82). Particularly, in these cases, the decoder can determine whether all CTUs in the picture are in a tile for which a tile size has been decoded.

Alternatively, the decoder can determine whether the tiles corresponding to the decoded tile sizes jointly cover the entire picture. Regardless of the particular manner in which the decoder decodes the tile sizes, the decoder configured according to this embodiment uses the decoded tile sizes to decode the encoded representation of the picture.

As seen in Figure 16B, an encoder can be configured according to this embodiment to encode a representation of the picture of a video bitstream using one or more of the following aspects. Particularly, as seen in method 90, the encoder can, in one embodiment, be configured to encode a specific spatial unit size to use for encoding the at least one picture (box 92). In one embodiment, the encoder can be configured to encode more than one tile size explicitly into the bitstream (box 94). In these cases, the encoder can be configured to encode each tile size by inserting one coded integer tile width value indicating the width of the units, and one coded integer tile height value into the bitstream indicating the height of the units. In one embodiment, the encoder can also be configured to derive at least one tile size by setting the tile size equal to a previously derived or decoded tile size (box 96). In one embodiment, the encoder can also be configured to insert a flag value for at least one current tile (box 98). In these embodiments, one value could be used to specify to the decoder that the tile size of the current tile should be copied from a previously decoded tile size, or the encoder could insert the tile size explicitly into the bitstream so that the decoder can derive the tile size explicitly. In one embodiment, the encoder can also be configured to indicate the number of tile sizes that are signaled by inserting a number value specifying the number of tile sizes into the bitstream. Regardless of the particular manner in which the encoder encodes the information, the encoder is configured to use the tile sizes to encode the at least one picture into the video bitstream.

Table 3 identifies an example syntax for use in this embodiment of the present disclosure. The syntax could be implemented on top of the currently existing HEVC video coding specification.

able 3 - Example syntax

tiles_enabled_flag equal to 1 specifies that there is more than one tile in each picture referring to the PPS. tiles_enabled_flag equal to 0 specifies that there is only one tile in each picture referring to the PPS.

It is a requirement of bitstream conformance that the value of tiles_enabled_flag shall be the same for all PPSs that are activated within a CVS. flexible_tiles_enabled_flag equal to 1 specifies that flexible tiles are used to partition pictures referring to the PPS into tiles. flexible_tiles_enabled_flag equal to 0 specifies that flexible tiles are not used to partition pictures referring to the PPS into tiles.

If flexible_tiles_enabled_flag is equal to 1 , the value of the variable tilejd and the value of the two-dimentional array CTU_tile_assigment are specified as follows: tilejd = 0

for( y= 0; y < PicHeightlnCtbsY; y++ )

for( x = 0; x < PicWidthlnCtbsY; x++ )

CTUJile_assignment[ x ][ y ] =“no tile” number_of_tilesJn_picture_minus2 plus 2 specifies the number of tiles in a picture.

number_ofJilesJn_picture_minus2 shall be in the range of 0 to PicSizelnCtbsY - 2, inclusive. use_previous_tile_size_flag equal to 1 specifies that the size of the current tile is equal to the size of the previous tile. use_previousJile_sizeJlag equal to 0 specifies that the size of the current tile is not equal to the size of the previous tile. When not present, the value of use_previousJile_sizeJlag is inferred to be equal to 0.

When use_previousJile_sizeJlag is equal to 1 , the value of the i’th element of the arrays tile_height_minus1 and tile_width_minus1 are derived as follows: tile_height_minus1 [ i ] = tile_height_minus1 [ i -1 ]

tile_width_minus1 [ i ] = tile_width_minus1 [ i -1 ] tile_width_minus1 [ i ] plus 1 specifies the width of the i-th tile in tile units.

tile_height_minus1 [ i ] plus 1 specifies the height of the i-th tile in tile units.

The value of the variable tilejd and the value of the two-dimentional array CTUJile_assigment are derived as follows: for( tile_coordinate_y = 0; tile_coordinate_y < PicHeightlnCtbsY; tile_coordinate_y++ ) for( tile_coordinate_x = 0; tile_coordinate_x < PicWidthlnCtbsY; tile_coordinate_x++ ) if( CTU_tile_assignment[ tile_coordinate_x ][ tile_coordinate_y ] ==“no tile” ) break

for( ctu_y = 0; ctu_y < ( tile_height_minus1 [ i ] + 1 ) && (tile_coordinate_y + ctu_y ) <

PicHeightlnCtbsY; ctu_y++ )

for( ctu_x = 0; ctu_x < ( tile_width_minus1 [ i ] + 1 ) && (tile_coordinate_x + ctu_x ) < PicWidthlnCtbsY; ctu_x++ )

CTU_tile_assignment[ tile_coordinate_x + ctu_x ][ tile_coordinate_y + ctu_y ] = tilejd

tile_id++

It is a requirement of bitstream conformance that all of the following conditions shall be true prior to the assignment of CTU_tile_assignment[ tile_coordinate_x + ctu_y ][ tile_coordinate_y + y ] to the value tilejd:

1. The value of tile_coordinate_y + ctu_y shall be less than PicHeightlnCtbsY

2. The value of tile_coordinate_x + ctu_x shall be less than PicWidthlnCtbsY

3. The value of CTUJile_assignment[ tile_coordinate_x + ctu_x ][ tile_coordinate_y +

ctu_y ] shall be equal to“no tile”

The derivation of the variable tilejd and the value of the two-dimensional array CTUJile_assigment as specified in the semantics of tile_height_minus1 [i ] are executed for each value of i. This means that the derivation is done after each loop of the variable i, including when use_previous tile_size lag is not present, when use_previousJile_sizeJlag is equal to 0, and when use_previousJile_sizeJlag is equal to 1. Likewise, the three bitstream conformance conditions shall be true for any assignment done in any loop.

Fourth Embodiment

In a fourth embodiment, the present disclosure provides a method for determining whether a given unit, such as unit U, does or does not begin a new tile. Particularly, the size of the units are defined and then scanned in a predefined order. For example, in one embodiment, the units is defined as being the same size as the CTU, and then scanned in raster scan order.

In raster scanning, scanning is conducted horizontally left-to-right at a steady rate, then blanks and rapidly moves back to the left, where it turns back on and sweeps out the next line such that the vertical position steadily increases. Of course, one of ordinary skill in the art will appreciate that when raster scan order is described as an example in this embodiment and other embodiments, other predefined scanning orders could be used (e.g., scanning horizontally right- to-left or scanning initially vertically). Regardless of the particular scanning order, for each unit, there is a flag specifying whether the unit starts a new tile or not. No flag is sent for a given unit when another unit positioned above and to the left belongs to the same tile. In such cases, it can be inferred that both the given unit and the other unit are positioned in the same tile.

Table 4 graphically illustrates whether the positioning of a unit being scanned indicates that the unit starts a new tile, and the values (if any) to set the flag indicating whether the unit starts a new tile.

Table 4: Some possible unit positioning and associated signaling For all cases shown in Table 4, except for the ambiguous case in the last row (i.e. , whether the unit being scanned is in tile B, tile C, or starts a new tile), one flag and the known status of other previously scanned units (i.e., the units positioned to the left and top of the unit being scanned, if they exist) is sufficient with which to specify the correct status of the unit currently being scanned. To remove the ambiguity in the case shown in the last row of Table 4, the present disclosure introduces a disambiguator bit. For example, in one embodiment, all the disambiguator bits in the defined scanning order are placed into a disambiguator array. The array is then explicitly signaled to the decoder.

Table 5 provides a first example syntax for use with this embodiment. The syntax shown in Table 5 could be implemented on top of the currently existing HEVC video coding specification. Additionally, for illustrative purposes only, the size of a unit is equal to the size of the CTU.

Table 5: First example syntax

Semantics for the Syntax Elements in Table 5

tiles_enabled_flag: This flag is set to 1 to specify that there is more than one tile in each picture referring to the PPS. This flag is set to 0 to specify that there is only one tile in each picture referring to the PPS.

tiles_status_flag_enabled This flag is set to 1 to specify that the tiling structure is signalled using the status flag method. This flag is set to 0 to specify that the tiling structure is not signalled using the tiles status flag method.

ctu_status[i] This is set to 0 to specify that the tile ID for the current CTU is copied from an available neighboring tile using copy_id(). This is set to 1 to specify that a new tile ID is set for the current CTU using set_new_id().

As seen in Table 4, there are some instances where the CTU statuses remain undetermined (e.g., those in rows 2, 3, 5, 6, and 8 of Table 4). In these cases, the status can be determined as specified by the signaling details seen in Table 6.

Table 6: Signaling Details for Ambiguous Cases

ctu_status_disambiguator_flag: This flag comprises the disambiguator bit that is used together with ctu_status to determine the status of the CTU when needed. The length of this bitstream equals the number of disambiguitor bits in the ctu_status. The

ctu_status_disambiguator_flag is set if and only if the value of the ctu_status_flag is 0. In these cases, when the disambiguator_bit = 0, the ctu_status should be copied from the unit on the left. When the disambiguator_bit = 1 , the ctu_status should be copied from the unit above. exist(): This function indicates whether a tile ID exists for a neighboring block with a tile ID (tileldAboveLeft, tileldAbove, or tileldLeft).

set_new_id(): This function sets a new tile ID for the current CTU.

copy_id(): This function copies the tile ID from a neighboring block according to the details provided in Table 6.

assign_ctu_to_tile(): This function assigns a given CTU to an explicit tile ID using the ctu_status and ctu_status_disambiguator flags. It also assigns the given CTU the status of a neighbor CTU positioned on the immediate left of the given CTU, immediately above the given CTU, and immediately above and to the left (i.e., diagonally) of the given CTU, if they exist. read_more_tiles( ): This function is read in a predefined scanning order, such as raster scan order, as follows:

IF there is at least one CTU in the picture that is not assigned to a tile, set the return value of read_more_tiles( ) to TRUE;

ELSE set the return value of read_more_tiles( ) to FALSE

In one aspect of this embodiment, the tile ID is signaled when a new tile is signaled. The tile ID can be a new tile ID or a previously defined tile ID. Signaling the tile ID in this manner allows for a more flexible tile structure with non-rectangular or discontinuous tiles.

Bit count Figures 17A-17C illustrate, according to one embodiment, the ctu_status bits (Figure 17B) for the tile partition (Figure 17A) of the picture seen in Figure 14. In this embodiment, cells that are empty indicate that no bit for that CTU is signaled in ctu_status. The“?” indicator signifies that a disambiguator bit is needed. Figure 17C illustrates the corresponding ctu_status disambiguator bits.

Table 7 provides a second example syntax for use with this embodiment. The syntax shown in Table 7 would replace the first example syntax seen in Table 5 and could be implemented on top of the currently existing HEVC video coding specification. Additionally, for illustrative purposes only, the size of a unit is equal to the size of the CTU.

Table 7: Second Example Syntax

Semantics for the Syntax Elements in Table 7

tiles_enabled_flag: This flag is set to 1 to indicate that multiple tiles in each picture refer to the PPS. This flag is set to 0 to indicate that only a single tile in each picture refers to the PPS. tile_flag: This flag indicates whether the tile ID of a current CTU is set equal to the tile ID of a neighboring CTU or starts a new tile.

When tile_flag is equal to 0, the value of CTU_tile_assignment[ x ][ y ] is specified as follows:

CTU_tile_assignment[ x ][ y ] = previous_tile;

When tile_flag is equal to 1 , the value of CTU_tile_assignment[ x ][ y ] and tilejd are specified as follows:

CTU_tile_assignment[ x ][ y] = tilejd;

tilejd = tilejd + 1 ;

tile_flag1 : This flag indicates whether the tile ID of a current CTU is set equal to the tile ID of a neighboring CTU or starts a new tile.

When tile lagl is set to 1 , the value of CTU ile_assignment[ x ][ y ] and tilejd are specified as follows:

CTUJile_assignment[ x ][ y ] = tilejd;

tilejd = tilejd + 1 ;

tile flag2: This flag indicates whether the tile ID of a current CTU is set to equal the tile ID of the CTU immediately above the current CTU, or immeditely to the left of the current CTU. When tile lag2 is set to 0, the value of CTU ile_assignment[ x ][ y ] is set to the value of the CTU to the left of the current CTU as follows:

CTUJile_assignment[ x ][ y ] = CTUJile_assignment[ x ][ y - 1 ]; When tile lag2 is set to 1 , the value of CTU ile_assignment[ x ][ y ] is set to the value of the CTU immediately above the current CTU as follows:

CTUJile_assignment[ x ][ y ] = CTUJile_assignment[ x - 1 ][ y ];

Fifth Embodiment In a fifth embodiment, the present disclosure provides a method for scanning the tiles similar to that provided by the first embodiment above. However, this fourth embodiment includes one or more of the following additional features. In particular, this embodiment also allows:

1. The use of traditional tiles in a“traditional tile mode” and the use of flexible tiles in a “flexible tile mode.” The determination of mode to operate in is signaled, for example, using a flag, such as a flexible_tiles_enabled_flag ; and/or

2. A particular size specification for a tile. For example, this embodiment could specify using the size of a latest tile for a current tile, or the size of a tile to the immediate the left or above the current tile. The specified size can be applied to the current tile or to all remaining tiles, for example, by signaling a mode value.

Figure 18A is a flow diagram illustrating a method 1 10 for constructing and signaling the tiles according to this embodiment. The method 1 10 is implemented by an encoder and beings with the encoder determining whether a picture has only one tile (box 1 12). If so, the encoder sets the tile_enabled_flag equal to 0 (box 1 14) before ending method 1 10. However, if the encoder determines that the picture has a plurality of tiles, the encoder creates an empty list of tiles (box 1 16). Then, while tiles remain and following a predefined scan order (box 1 18), the encoder sets the copy_tile_size_mode (box 120). If the copy_tile_size_mode is set to 0 (box 122), the encoder adds the tile, including the tile height and width, to the list of tiles (box 124). The process then ends.

Regarding the decoder, it parses and constructs the tiles sent by the encoder. In this embodiment, the decoder can perform these functions according to the method 130 illustrated in Figures 18B-18C. As seen in Figure 18B, the decoder determines whether the

tile_enabled_flag is set to 0 (box 132). If so, the decoder will know that the picture comprises only a single tile, and therefore, method 130 ends. Otherwise, the decoder determines whether the flexible_tiles_enabled_flag was set to 1 (box 134). If not, the decoder performs its decoding functions using traditional tiles (box 136). If the decoder determines that the

flexible_tiles_enabled_flag was not set to 1 , however, the decoder begins processing the tiles.

In particular, while tiles remain to be processed (box 138), the decoder parses the

copy_tiles_size_mode value (bpx 140) and uses that value to determine the width and height of the current tile (i.e., the tile currently being processed).

Particularly, if copy_tile_size_mode is equal to 0 (box 142), the decoder parses the next tile in the tile list that was created by the encoder, specified by its width and length following the predefined scan order (box 144).

If copy_tile_size_mode is equal to 1 (box 146), the decoder copies the width and height of the last signaled tile to the width and height of the current tile (box 148). Turning to Figure 18C, if copy_tile_size_mode is equal to 2 (box 150), the decoder copies the width and height of the left tile to the width and height of the current tile (box 152).

If copy_tile_size_mode is equal to 3 (box 154), the decoder copies the width and height of the tile immediately above the tile currently being processed to the width and height of the tile currently being processed (box 156).

If copy_tile_size_mode is equal to 4, and while not at the end of the list of tiles (box 158), the decoder copies the width and height of the last signaled tile to the width and height of the current tile (box 160).

If copy_tile_size_mode is equal to 5, and while not at the end of the list of tiles (box 162), the decoder copies the width and height of the tile to the immediate left of the current tile, and copies those values to the width and height of the current tile (box 164).

If copy_tile_size_mode is equal to 6, and while not at the end of the list of tiles (box 166), the decoder copies the width and height of the tile immediately above the current tile to the width and height of the current tile (box 168).

Table 8 provides example syntax for use with this embodiment. The syntax shown in

Table 8 could be implemented on top of the currently existing HEVC video coding specification.

Table 8: Example Tile syntax

Semantics for the Syntax Elements in Table 8

tiles_enabled_flag: When this flag is 1 , it indicates that there multiple tiles in each picture referring to the PPS. When this flag is 0, it indicates that each picture referring to the PPS has only a single tile. flexible_tiles_enabled_flag: When this flag is 1 , it indicates that flexible tiles are defined.

When this flag is 0, it indicates that the tiles are defined without using flexible tiles syntax.

copy_tile_si ^z e_mode: This value indicates how the decision about the current tile size should be made. Specifically:

• A copy_tile_size_mode equal to 0 indicates that the tile size is explicitly signaled

• A copy_tile_size_mode equal to 1 indicates that the last signaled size is used

• A copy_tile_size_mode equal to 2 indicates that the tile size to the left of the top/left sample of the current tile is used

• A copy_tile_size_mode equal to 3 indicates that the tile size above the top/left sample of the current tile is used

• A copy_tile_size_mode equal to 4 indicates that the last signaled tile is used for all remaining tiles

• A copy_tile_size_mode equal to 5 indicates that the tile to the left of the top/left sample of the current tile is used for all remaining tiles

• A copy_tile_size_mode equal to 6 indicates that the tile above the top/left sample of the current tile is used for all remaining tiles.

copy_f ^r o ^m _last_signaled_tile(): This is a function that copies the width and length of the last signaled tile to the width and length of the current tile.

copy_f ^r Om_above_tile(): This is a function that copies the width and length of the tile above the current tile to the width and length of the current tile.

This is a function that copies the width and length of the tile to the left of the current tile to the width and length of the current tile.

Sixth Embodiment

In a sixth embodiment, the present disclosure provides a method for partitioning a picture into units, and then scanning the units in a pre-defined scanning order. During this process, a tile identifier (tile id) is specified for each unit. The pre-defined order is in raster scan order and the unit size is equal to the size of one CTU unit.

Figures 19A and 19B are flow diagrams illustrating methods implemented by an encoder and a decoder, respectively, to construct and signal tiles. Particularly, method 170 seen in Figure 19A provides a method 170 by which the encoder constructs and signals the tiles. As seen in method 170, for each unit in the picture (box 172), the encoder signals a tile ID for the unit in the bitstream (box 174), and assigns the tile ID to the unit (box 176).

Method 180 seen in Figure 19B provides a method 180 implemented at the decoder. Particularly, for each unit in the picture (box 182), the decoder parses the tile ID for the tile with which the unit is associated (box 184), and assigns the tile ID to the unit.

An example syntax table and semantics description for embodiment 5 on top of the HEVC video coding specification may look as Table 9.

Table 9: Example Tile Syntax

Semantics for the Syntax Elements in Table 9

tiles_enabled_flag: When this flag is set to 1 , it indicates that there are more than one tile in each picture referring to the PPS. When this flag is set to 0, it specifies that there is only one tile in each picture referring to the PPS.

tile_id[ y ][x ]: This specifies the tile ID of the CTU at coordinates y and x. The value of CTU_tile_assignment is specified as follows:

CTU_tile_assignment [ y ][ x ] = tile_id[ y ][ x ].

Seventh Embodiment

This embodiment of the present disclosure is similar to that of the fifth embodiment, but includes the following additional features.

First, a flag is set to determine whether flexible tiles are enabled, or whether a traditional tile structure, such as the current HEVC tile structure, is deployed.

Second, this embodiment defines a method for coding independent tiles using a fixed code value, such as code value equal to 0. A tile is defined to be independent if it has the same size as the unit (e.g. CTU). This feature beneficially saves bits in cases where a given picture comprises a plurality of independent tiles.

Third, this embodiment defines a method that allows for the optional encoding of tile ids with fixed length coding. In such cases, a flag is used to signal whether fixed length coding of tile ID values is used, or whether variable length coding of tile ID values is used.

Those of ordinary skill in the art will appreciate that the present disclosure is not limited to these features and embodiments. Rather, it is possible to vary these features and how they are combined in various embodiments. Further, not all features need to be implemented. For example, using a flag to indicate flexible tiles and the method of coding independent tiles could be used together even though the capability to encode tile ids with fixed length coding is not available. Table 10 provides example syntax for use with this embodiment. The syntax shown in Table 10 could be implemented on top of the currently existing HEVC video coding specification.

able 10: Example Tile Syntax

tiles_enabled_flag: When this flag is set to 1 , it indicates that there is more than one tile in each picture referring to the PPS. When this flag is set to 0, it indicates that there is only one tile in each picture referring to the PPS.

flexible_tiles_enabled_flag: When this flag is set to 1 , it indicates that flexible tiles are defined. When this flag is set to 0, it indicates that tiles are defined without using flexible tiles syntax.

use_zero_to_indicate_independent_tile_flag: When this flag is set to 1 , it indicates that independent tiles are coded with tilejd equal to 0. An independent tile is a tile with a tile width equal to CTU width, and a tile height equal to CTU height. When this flag is set to 0, it indicates that independent tiles are coded in the same way as dependent tiles. use_fixed_length_code_for_tile_id_flag: When this flag is set to 1 , it indicates that tile ids are encoded using fixed length coding. When this flag is set to 0, it indicates that tile ids are encoded using fixed length coding.

number_of_unique_tile_id: This is the value to use for the tile ID to indicate that a tile depends from all other tiles.

tile_id[ i ][ j ] : This value specifies the tile ID of the CTU with coordinates i and j.

Eighth Embodiment

In an eighth embodiment of the present disclosure, the picture is divided into tiles using quad trees and/or flexible tile splitting. The process for dividing a picture using quad trees in this embodiment is similar to the HEVC process for splitting CUs.

For simplicity in explaining this embodiment, the smallest tile unit is set to the CTU size. However, those of ordinary skill in the art should readily appreciate that for other embodiments of the present disclosure, the smallest tile unit is a CU or a rectangular set of CTUs.

Additionally, not all tiles have a width and height that are a factor of 2 in terms of CTUs.

In one embodiment, the splits for the quad trees use integer division if the height and width are not a factor of 2. For example, if the size of a given tile is 5x7 CTUs, a quad tree split would divide the tile into tiles T-i, T ₂ , T3, and T4 of size 2x3, 2x4, 3x3 and 3x4, respectively, such as that illustrated in Figure 20. Similarly, a horizontal or vertical“midway” split also uses integer divisions.

A“midway” split is defined herein such that for a given tile A having length N CTUs, where N is a positive integer larger than 1 , the resulting first split tile B has the length N/2 and the second split tile C has the length N/2 + N%2 where 7’ indicates integer division and‘%’ denotes a modulo operator. The length of a tile could be either the width of the tile, the height of the tile or both. For example, if a given tile has a size 1x5, that tile can be split, according to

embodiments of the present disclosure, into two tiles - a first tile having a size of 1x2, and a second tile having a size of 1x3.

Figures 21A-21 B illustrate a method 190 implemented by an encoder to construct and signal tiles. As seen in Figure 21 A, the encoder first sets a picture/slice as the only tile in a list of uncompleted tiles (box 192). The encoder then creates an empty list of completed tiles (box 194). The encoder then checks to determine whether any uncomplete tiles are available (box 196). If not, method 190 ends. Otherwise, the encoder obtains the next uncomplete tile in raster scan order (box 198). The encoder then determines whether the tile could be further split (box 200). If not, the encoder removes the tile from the list of uncompleted tiles (box 202), and adds the tile including tile ID, tile position and tile dimensions to the list of completed tiles (box 204) before returning to determine whether any other uncompleted tiles are available (box 196). The encoder then signals the flag indicating whether the tile should be split according to a quad tree split (or a“midway” horizontal or vertical split if the height or width equals the height or width of the CTU) (box 206). If the flag does indicate a split, the encoder removes the current tile from the list of uncompleted tiles (box 208), and adds the split tiles to the list of uncompleted tiles (box 210) before returning to determine whether any other uncompleted tiles are available (box 196).

As seen in Figure 21 B, if the flag does not indicate a split, the encoder signals a flag indicating whether the tile should be further split (box 212). If the tile is not to be split, the encoder again determines whether any other uncompleted tiles are available (box 196). If the tile should be further split, however, the encoder signals a flag indicating whether the tile should be split vertically or horizontally (box 214). Note, however, that the vertical/horizontal split indication is not necessary when the tile width equals the CTU width or the tile height equals the CTU height.

The encoder then signals the particular CTU position at which to split the tile (box 216) and splits the tile at that position (box 218). So split, the encoder removes the current tile from the list of uncompleted tiles, and adds the split tiles to the list of uncompleted tiles (box 220) before returning to determine whether any other uncompleted tiles are available (box 196).

Figures 21 C-21 D illustrate a method 230 implemented by a decoder to parse the tiles constructed and signaled by the encoder in Figures 21A-21 B. As seen in Figure 21 C, the decoder first sets a picture/slice as the only tile in a list of uncompleted tiles (box 232). The decoder then creates an empty list of completed tiles (box 234). The decoder then checks to determine whether any uncomplete tiles are available (box 236). If not, method 230 ends. Otherwise, the decoder obtains the next uncomplete tile in raster scan order (box 238). The decoder then determines whether the tile could be further split (box 240). If not, the decoder removes the tile from the list of uncompleted tiles (box 242), and adds the tile including tile ID, tile position and tile dimensions to the list of completed tiles (box 244) before returning to determine whether any other uncompleted tiles are available (box 196). If the tile could be further split, the decoder then parses the flag indicating whether the tile should be split according to a quad tree split (or a“midway” horizontal or vertical split if the height or width equals the height or width of the CTU) (box 246). If the flag does indicate such a split, the decoder removes the current tile from the list of uncompleted tiles (box 248), and adds the split tiles to the list of uncompleted tiles (box 250) before returning to determine whether any other uncompleted tiles are available (box 236).

As seen in Figure 21 D, the decoder parses a flag signaled by the encoder to determine whether the tile should be further split (box 252). If the tile is not to be further split, the decoder returns to determine whether any other uncompleted tiles are available (box 236). If the tile should be further split, however, the decoder parses a flag indicating whether the tile should be split vertically or horizontally (box 254). Note, however, that parsing this flag to determine a vertical/horizontal split is not necessary when the tile width equals the CTU width or the tile height equals the CTU height.

The decoder then parses a code word to determine the particular CTU position at which to split the tile (box 256) and splits the tile at that position (box 258) before returning to determine whether any other uncompleted tiles are available (box 236).

Table 1 1 provides example syntax for use with this embodiment. The syntax shown in Table 1 1 could be implemented on top of the currently existing HEVC video coding

specification.

Table 11 : Example Tile Syntax

Semantics for the Syntax Elements in Table 11

tiles_enabled_flag: This flag is set to 1 to indicate that there are more than one tile in each picture referring to the PPS. This flag is set to 0 to indicate that there is only one tile in each picture referring to the PPS.

quadtree_split_flag: This flag is set to 1 to indicate that the tile is split using quadtree if tile_width is larger than ctu_width and tile_height is larger than ctu_height, split horizontally if tile_width equals ctu_width and tile_height is larger than ctu_height or split vertically if tile_width is larger than ctu_width and tile_height equals ctu_height. This flag is set to 0 to indicate that the tile is not split using quadtree split. Note that bitstream conformance requires that the value of quadtree_split_flag shall be equal to 0 when tile_width is equal to ctu_width, and tile_height is equal to ctu_height.

flexible split flag: This flag is set to 1 to indicate that the tile is split using flexible tile split. This flag is set to 0 to indicate that the tile is no further split.

split_in_horizontal_direction_flag: This flag is set to 1 to indicate that the flexible split is made in horizontal direction. This flag is set to 0 to indicate that the flexible split is made in vertical direction. If the split_in_horizontal_direction_flag is not present, the flag value is inferred to be 1 if tile_width equals ctu_width and tile_height is larger than ctu_height, and inferred to be 0 if tile_width is larger than ctu_width, and tile_height equals ctu_height.

split_position_minus1 plus 1 specifies the split position for the flexible split in units of CTUs. split_position_minus1 should be in the range from 0 to tile_width minus 2 if

split_in_horizontal_direction_flag is set to 1 , or in the range from 0 to tile_height minus 2 if split_in_horizontal_direction_flag is equal to 0.

read next_tile(): This function reads tiles recursively, and moves to the next tile in raster scan order if there is no more splits to be made for the current tile. No further split is required if either of the following cases apply.

1. The tile width equals the CTU width and the tile height equals the CTU height; and/or 2. The flexible_split_flag has been set to 0.

When it is determined that a tile is not to be split further, the tile including tile ID, its x and y positions, width and height, are added to a list of completed tiles.

split_tile_horizontally(): This function splits the current tile horizontally into tileA and tileB at the position tile_height/2 in terms of CTU height where the division is an integer division.

split_tile_vertically(): This function splits the current tile vertically into tileA and tileB at the position tile_width/2 in units of CTU width, where the division is an integer division.

split_tile_using_quadtree(): This function splits the current tile horizontally and vertically into tileA, tileB, tileC, and tileD at the positions tile_height/2 in terms of CTU height and tile_width/2 in terms of CTU width where the divisions are integer division.

split_tile_using_flexible_split(): This function splits the current tile horizontally if

split_in_horizontal_direction equals 1 , or vertically if split in horizontal_direction_flag equials 0 into tileA and tileB at position split_position_minus1 plus 1.

According to embodiments of the present disclosure, the same tile structure can be expressed with different variations of the syntax. That is, splits can be ordered differently but still achieve the same result.

Additionally, it is often possible to assign a split_position_minus1 value that is higher than the allowed value. For example, if the tile_width is 6, then 3 bits are needed to select any of the 5 split positions of the tile. Thus in this case, 2 ³ - 5 = 3 positions cannot be selected although it could be possible to select them without the constraint in the semantics. In one embodiment, where the number of positions is greater than 2, the additional positions are used to signal the next flags in the parsing. If at least 2 positions are available the next

quadtree_split_flag, if available, is set to 0 if split_position_minus1 +1 = tile_width if

split_in_horizontal_direction equals 0 or split_position_minus1 +1 = tile_height if

split_in_horizontal_direction equals 1. The next quadtree_split_flag is set to 1 if

split_position_minus1 +1 = tile_width +1 if split_in_horizontal_direction equals 0 or

split_position_minus1 +1 = tile_height +1 if split_in_horizontal_direction equals 1.

In this embodiment, the quadtree_split_flag is not set to 1 in the tile structure of a picture like that of Figure 14, meaning that every tile split termination costs 2 bits. In some

embodiments, the order of the syntax element could be different than what is provided in the proposed tile syntax. For example, the flexible_split_flag could be put before the

quadtree_split_flag, where a quad_tree_split_flag equal to 0 would terminate further splitting of the tile. In this example the tile structure of Figure 15 could be expressed using 19+10 =29 bits and the tile structure of Figure 14 could be expressed using 73-19 = 54 bits. In one embodiment an extra flag could be signaled before parsing the tile structure indicating whether

quad_tree_split_flag or flexible_split_flag should be parsed first. In yet another embodiment, the quadtree split_flag is not used. Rather, only the flexible_split_flag is used. In this aspect, after the split_in_horizontal_direction_flag, a flag“split_midways_flag” could be used to specify whether the tile should be split midways or if position split_position_minus1 should be read to do the split at that position.

Ninth Embodiment

In this embodiment, the tile structure is described by specifying the boundaries between the tiles. Within this disclosure, the boundaries between tiles are also referred to as“divider lines.”

In one aspect of this embodiment, the divider lines are straight lines described by their start and end points. Since the start and end points are interchangeable without affecting the divider line, both are referred to hereafter as end points.

As the divider lines are either horizontal or vertical, three coordinates and a direction flag are sufficient with which to uniquely describe each line. Further, this embodiment uses one flag for signaling the group of horizontal divider lines, and one flag for signaling the group of vertical divider lines. This further reduces the total bit count.

For example, the example tile structure illustrated in Figure 14 can be expressed using 9 horizontal (hi , h _å , !¾, tu, h ₅ , he,) and vertical (v-i, v _å , V ₃ ) divider lines, illustrated in Figures 22A- 22C. This equates to 3 ^* 9=36 code words for the coordinate of the end points, and 2 flags for the groups of horizontal and vertical divider lines, respectively.

Figure 23A is a flow diagram illustrating a method 260 implemented at the encoder for determining the horizontal and vertical divider lines for a given tile structure. For illustrative purposes only, method 260 is described in the context of the tiling structure for the picture seen in Figure 14.

As seen in Figure 23A, the encoder determines the value of the tile_enabled_flag (box 262). If this flag is set to 0, it means that the picture comprises only one tile having the same width and height as the picture. Thus, the process ends.

If this flag is not set to 0, however, the encoder creates a list of all horizontal divider lines (AHDL) that form the desired tiling structure (box 264), as well as a list of all vertical divider lines (AVDL) that form the desired tiling structure (box 266). If the AHDL is not empty (box 268), the encoder signals horizontal divider lines and sends AHDL to the decoder (box 270). Similarly, if the AVDL is not empty (box 272), the encoder signals the vertical divider lines and sends the AVDL to the decoder (box 274). The order of creating and signaling AHDL and AVDL may be reversed without affecting the general aspects of the embodiment.

Figure 23B is a flow diagram illustrating a method 280 implemented at the decoder for processing tiles having a given tile structure described with corresponding horizontal and vertical divider lines. As above, this embodiment of the decoder functionality is described in the context of the tiling structure for the picture seen in Figure 14. However, this is for ease of description only. As seen in Figure 23B, the decoder first determines whether the tile_enabled_flag is set to 0 (box 282). If this flag is set to 0, it means that the picture comprises only one tile having the same width and height as the picture. Thus, the process ends.

If the flag is not set to 0, however, the decoder creates an empty complete_tile_list (box 284), parses the AHDL, and creates a list of horizontal_ divider Jines (HDL) (box 286). The decoder also parses the AVDL, and create a list of vertical_ divider Jines (VDL) (box 288). The order of parsing AHDL and AVDL may be reversed without affecting the general aspects of the embodiment.

The decoder then creates a cornerjist with all the corners and their directions created by picture boundaries, HDL and VDL (box 290). For example, referring to the illustration of Figure 22, a cross point“- -” specifies four separate corners in the cornerjist, and a T-junction “ {" or“-|” or“ _T ” or“- ¹ -” specifies two separate corner points.

Once the cornerjist has been created and populated, the decoder processes that list. Particularly, as long as the cornerjist is not empty (box 292), the decoder obtains one or more corners in the predefined scan order. In particular the decoder picks a first corner in the scan order and locates the three corresponding corners which form a tile from the cornerjist (box 294). The decoder then adds the tile to the complete_tile_list (box 296) and removes the four used corners from the cornerjist (box 298). This process continues until the cornerjist is empty.

Table 12 provides example syntax for use with this embodiment. The syntax shown in

Table 12 could be implemented on top of the currently existing HEVC video coding

specification. Each end point in the horizontal/vertical list of divider lines, which is not located on the picture boundaries, should be located on the length of one divider line in the

vertical/horizontal list (i.e., not another start or end point). This will help guaranty proper tile partitioning.

Table 12: Example Tile Syntax

Semantics for the Syntax Elements in Table 12

tiles_enabled_flag: This flag is set to 1 to indicate that there are more than one tile in each picture referring to the PPS. This flag is set to 0 to indicate that there is only one tile in each picture referring to the PPS.

tile_divider_line_enabled_flag: This flag is set to 1 to indicate that divider lines are used to describe the tiling structure. This flag is set to 0 to indicate that divider lines are not used to describe the tiling structure.

horizontal- divider _lines_count: This value specifies the number of horizontal divider lines vertical- divider _lines_count: This value specifies the number of vertical divider lines horizontal- divider _lines[ i ]: This value lists all the horizontal divider lines. Divider lines is specified by the coordinates of their ends. Because every divider line in this list is horizontal, three coordinates are enough to describe each divider line uniquely - i.e. , a single y value and two x values.

vertical- divider _lines[ i ]: This value lists all the vertical divider lines. Divider lines is specified by the coordinates of their ends. Because every divider line in this list is vertical, three coordinates are enough with which to describe each divider line uniquely. - i.e., two y values and one x value.

corner_list: This is the list of all corner points described by their position and direction ( _G , , ^L , ^J ). This list considers all the picture borders and all the horizontal and vertical divider lines (including their crossings).

In one aspect, this embodiment of the disclosure assumes a full grid. The start and end points of the divider lines that should be removed are then specified. To describe the tiling structure shown in Figure 15, this embodiment describes 6 divider lines that should be removed, and thus, 3 ^* 6 code words plus two code words for the group of horizontal and vertical divider lines are required to describe the tiling structure.

In yet another aspect, the divider lines have bending point(s) in addition to start and end points. Bend points represent“joints.” According to the present disclosure, joints function as an end point to a previous point, and as a start point for the next point in the line description.

Multiple sets of divider lines can result in the same tiling structure. The coder can optimize for the minimum number of divider lines. In one aspect, the divider lines can have parts in common. In another aspect, the divider lines do not have parts in common.

In yet another aspect, the divider lines is specified once and then copied to given position in the grid. Tenth Embodiment

This embodiment of the present disclosure describes the tile structure using a set of points that are in the structure and that have specific properties. These points are referred to herein as“T-junctions.”

As previously described, the tile structure comprises a set of internal lines, also referred to as“divider lines.” Each divider line is a straight line with two end points where the divider line terminates (i.e., the divider line is not further continued). The so-called“T-junctions” are defined at the position of these end points of the divider lines. Because each divider line ends at the junction of that divider line with another perpendicular divider line (not an end point of another divider line because of rectangular tiling structure), or at the junction with a border of the picture perpendicular to the first divider line, each T-junction has one of the four possible shapes: - ¹ -, y, -| , |-. Hence each T-junction can be fully described using the position of the junction and one of the 4 shapes -p, -| , |- . The three directions pointed at by each T-junction are referred to herein as the“arms” of that T-junction.

The tiling structure of the picture can then be determined from the set of the T-junctions by extending all three arms of all T-junctions in the picture until each arm reaches either the border of the picture or another T-junction position.

In one aspect of this embodiment, the set of T-junctions is shortened by removing redundant T-junctions. As described herein, redundant T-junctions are those which are located on the boundary of the picture and do not change the tiling structure if they are removed. An example of a tiling structure with redundant T-junctions is seen in Figures 24A-24B. The set of T-junctions T-J after removing the redundant T-junctions is also presented in Figure 24C. Redundancy is checked on a one by one basis. This means that if a T-junction T-J is redundant, and thus, will be removed from the set, it should not be included in redundancy check for the rest of the T-junctions T-J in the set.

Figures 25A-25B illustrate the example tiling structure S and a corresponding set of T- junctions T-J that describes that specific tiling structure S.

Figure 26A is a flow diagram illustrating a method 300 implemented at the encoder for creating the set of T-Junctions as previously described. Method 300 begins with the encoder selecting a tile structure and creating a set of divider lines that correspond to that tile structure (box 302). The encoder then creates the set of T-junctions (TJ) (box 304). In this embodiment the STJ includes information regarding the position and orientation of each T-junction. This can be accomplished using a variety of processes, however, in this embodiment; the encoder assigns one T-junction to the position of each end of a divider line in the created set, and assigns the proper direction to that T-junction based on whether the divider line is horizontal or vertical, and whether the end point is a left/right or a top/down end point. The encoder then removes redundant T-junctions. Particularly, for each T-junction on the picture boundary (box 306), the encoder determines whether there is another T-Junction in the STJ with the same column or row number, and if so, whether the direction different from the T-junction being examined (box 308). If so, the encoder removes that T-junction (box 310).

Figure 26B is a flow diagram illustrating a method 320 implemented at the decoder for creating the set of T-Junctions as previously described. Method 300 begins with the decoder decoding a set or list of T-junctions (box 322). Each T-junction has a position and a direction or type. There are 4 directions/types. No two T-junctions should have the same position and no T-junction should have a position outside the picture. Derive the tile partition structure Then, the decoder derives the tile partition from the set or list of T-junctions such that each T-junction is extended from all its three arms until each arm reaches another T-junction or the picture boundary (box 324).

Table 13 provides example syntax for use with this embodiment. The syntax shown in Table 13 could be implemented on top of the currently existing HEVC video coding

specification.

Table 13: Example Tile Syntax

Semantics for the Syntax Elements in Table 13

tiles_enabled_flag: This flag is set to 1 to indicate that there are more than one tile in each picture referring to the PPS. This flag is set to 0 to indicate that there is only one tile in each picture referring to the PPS. tiles_enabled_flag: This flag is set to 1 to indicate that there are more than one tile in each picture referring to the PPS. This flag is set to 0 to indicate that there is only one tile in each picture referring to the PPS.

tile_tJu ⁿ ctio ⁿ _enabled_flag: This flag is set to 1 to indicate that the T-junctions are used to describe the tiling structure. tJunction_size_minus1 plus one: This value specifies the number of T-junctions being signaled.

tJunction_new_row_seq: This value is a bit sequence with the length equal to the number of T-junctions.

tJunction_new_row_seq: This value starts at 1 and toggles every time a T-junction changes the scan line. As an exemplary case, for the tile structure of 14, the t Junction_new_row_seq is set to: 1 1 100101 10. Using this, we do not need to signal both x and y for a T-junction as long as it is located at the same scan line as the previous T-junction.

tjunctionjist: This is the list of all T-junctions (including their position and direction) in the defined scan order.

complete_tile_list: This is a list which keeps all the completed tiles. Completed tiles is specified with their four corners. complete_tile_list is initialized with one tile including all units in the picture.

update_complete_tile_list(): This is defined as a function that updates the list of completed tiles and it is described as follows:

update_complete_tile_list() {

• Using the position and direction of tjunctionjist [i], extend all arms of the tjunctionjist [i] until each arm reaches another T-junction or the picture boundary.

• if the units on the two sides of each arm are assigned to the same tile, update the

complete tile list by adding the newly created tile(s).

}

Eleventh Embodiment

This embodiment is based on the observation that the number of tiles is usually much lower than the number of blocks in the picture. This embodiment exploits this sparsity for coding the tiling structure.

In more detail, each tile can be uniquely specified with the position of its two opposite corners. For simplicity, the top left corner is considered as the start position of the tile, and the bottom right corner is considered as the end position of the tile. Corresponding start and end point coordinates of the tiles are marked in a matrix with the number of columns and rows equal to the picture width and height divided by the width and height of the picture units (e.g. CTUs). The matrix corresponding to start corner coordinates on the grid are considered to be a sparse matrix and are compressed using sparse matrix compression methods, such as compressed sparse column (CSC) or compressed sparse row (CSR), for example. The same applies to the end corner matrix. Another possible way to compress the sparse matrix is to compress the matrix in the same or similar way as it is done in HEVC for the matrices of coefficient significance or coefficient signs when compressing transform coefficients. Figures 27A-27C show how the tiling structure of a picture 10 (Figure 27A) is related to the matrix of start corners (Figure 27B) and the matrix of end corners (Figure 27C).

One difference from the standard CSC and CSR is that the only non-zero values in the start corner matrix and end corner matrix are equal to 1. Therefore, what is sent in the bitstream is only the position of those non-zero values.

Figure 28A is a flow diagram illustrating a method 330 implemented at the encoder for creating the start and end corner matrices, as previously described. As seen in Figure 28A, method 330 begins with the encoder determining whether the picture has only one tile (box 332). If so, the encoder sets the tile_enabled_flag equal to 0 (box 334) and the process ends.

If the encoder determines that the picture has more than one tile, the encoder creates the start corners matrix and the end corners matrix by marking the start point and end point of each tile in the corresponding matrix (box 336). The encoder then creates the start corners bitstream and end corner bitstreams (e.g. using CSC or CSR) (box 338), and signals the compression type (CSC or CSR) by setting the sparse_matrix_compression_flag (box 340).

Figure 28B is a flow diagram illustrating a method 350 implemented at the decoder for processing the start and end corner matrices, as previously described. As seen in Figure 28B, method 350 begins with the decoder determining whether the tile_enabled_flag is set to 0 (box 352). If so, the process ends because the picture includes only a single tile, which has the same width and height as the picture.

If there is more than one tile in the picture, however, the decoder creates an empty list of tiles (box 354). The decoder then parses the compressed start corners and end corners from the bitstream, and expands them to the start corners matrix and end corners matrix (box 356). Then, provided that there are start corners defined in the start corner matrix (box 358), the decoder reads the position of the next start corner in the raster scan direction (box 360).

The decoder then locates the corresponding end corner in the end corner matrix, and adds the tile (e.g. tile position and tile id, or e.g. tile position and width and height) to the complete_tile_list (box 362). The decoder then sets the corresponding start and end corners in the matrix to zero and repeats the process.

As a sanity check, the decoder can determine:

• A number of start corners (ones in the start corner matrix) is equal to the number of tiles. • A number of end corners (ones in the start corner matrix) is equal to the number of tiles.

• A first element of the start corner matrix is always set to 1

• A last element of the end corner matrix is always set to 1

Table 14 provides example syntax for use with this embodiment. The syntax shown in Table 14 could be implemented on top of the currently existing HEVC video coding

specification.

able 14: Example Tile Syntax

Semantics for the Syntax Elements in Table 14

tiles_enabled_flag: This flag is set to 1 to indicate that there is more than one tile in each picture referring to the PPS. : This flag is set to 0 to indicate that there is only one tile in each picture referring to the PPS. sparse_matrix_compression_enabled_flag : This flag is set to 1 to indicate that the sparse matrix is used to describe the tiling structure.

sparse_matrix_compression_mode: This indicates which sparse matrix compression method is used for the start and end corner matrix. Setting this mode to 0 indicates that the compressed sparse row (CSR) method is used.

start_corners_stream_1 is the first array that is being used to build the start_corners_matrix. This array is constructed as: start_corners_stream_1 [i] = number of nonzero elements on the (i)-th row in the start_corners_matrix.

tiles_count: This value specifies the number of tiles. This also gives the number of non-zero values in the start_corner_matrix and end_corner_matrix.

start_corners_stream_2 is the second array that is being used to build the

start_corners_matrix. This array contains the column index (0 to n-1 ) of each element of start_corners_matrix and hence is of length number of non-zero elements in the start_corners_matrix.

end_corners_stream_1 is the first array that is being used to build the end_corners_matrix. This array is constructed as: end_corners_stream_1 [i] = number of nonzero elements on the (i)-th row in the end_corners_matrix.

end_corners_stream_2 is the second array that is being used to build the

end_corners_matrix. This array contains the column index (0 to n-1 ) of each element of end_corners_matrix and hence is of length number of non-zero elements in the

end_corners_matrix.

start_corners_matrix is the matrix of start corners which is uniquely defined by two arrays start_corners_stream1 and start_corners_stream_2.

end_corners_matrix is the matrix of end corners which is uniquely defined by two arrays end_corners_stream1 and end_corners_stream_2.

complete_tile_list is the list of complete tiles, initialized as an empty list.

update_complete_tile_list() is a function that reads tiles from start_corners_matrix and end_corners_matrix and adds them to the complete_tile_list. Assuming the raster scan direction left to right and top to bottom, the function is described as follows:

update_complete_tile_list() {

Find the first 1 value in the start_corners_matrix in the raster scan direction

Find the corresponding 1 value in the end_corners_matrix.

Add the tile to the complete_tile_list using the coordinates of the start corner and end corner.

Set the corresponding start corner to zero in the start_corners_matrix.

} Embodiments herein are described with reference to a single partition structure defined by an encoder or derived by a decoder, such as an encoder 400 and decoder 450 illustrated in Figures 34A and 34B. Figure 34A shows an example encoder 400 transmitting a bitstream 700, with information indicating a single partition structure 750. The partition structure 750 includes segments or tiles 720, 730 and 740. The bitstream 700 comprises information indicating a size of a unit 710, and information related to segments 720, 730 and 740, where an edge of segment 720 borders an edge of segment 730 and an edge of segment 740.

Figure 34B shows a corresponding example decoder receiving a bitstream 700, with information indicating a single partition structure 750. The partition structure 750 includes segments or tiles 720, 730 and 740. The bitstream 700 comprises information indicating a size of a unit 710, and information related to tiles or segments 720, 730 and 740, where an edge of segment 720 borders an edge of segment 730 and an edge of segment 740.

Of course, one of ordinary skill in the art will appreciate that an encoder 400 and decoder 450 can be used separately as shown in Figures 34A-34B or in conjunction to exchange a series of images as shown in Figure 29.

In such cases where a series of images are exchanged, the bitstream in one or more embodiments includes an indicator of the identity of the picture or otherwise associate a particular partition structure with a picture or multiple pictures. As shown in Figure 29, partition structure 400 shows divider lines 420 for picture segments 421 , 422, 423. A given picture segment of the partition structure comprises at least two subsegments, e.g., subsegments formed by a divider line 410 forming the border of a subsegment. A subsegment can comprise a plurality of units. For instance as shown in Figure 29, subsegment 480 comprises 4 units. An edge of a first picture segment 421 borders corresponding edges of adjacent picture segments 422 and 423. As shown in Figure 29, the encoder 400 and decoder 450 encode and decode respectively a plurality of images which have the same partition structure 400 or different partition structures 430 and 450. These different partition structures could be flexible partition structures as described herein. Alternatively or additionally as shown in Figure 29, the different partition structures can include as well traditional partition structures with divider lines 440 and 460 that go across the image.

Figure 35 shows an example system or method 800 in which an encoder or decoder described herein would be useful. The system comprises a media source (e.g., a capture, renderer or file). The media source 810 is fed to a projection mapping 820 and then to a tile encoder or transcoder, e.g. an encoder 400 for partitioning an image of the media source as described herein. The encoded information is stored in storage 830 and then fed to a server 840. It can be transmitted over a network 850 to a viewing device 860. For example, the image is one of video, and the viewing device 860 has a decoder 450 as described herein for decoding the partitioned image, rendering 870 the image and displaying the image on a display 880 to a user of the viewing device 860. Such a viewing device can be, for example, a Head Mounted Display (HMD) device or any display device.

One or more embodiments are applicable to applications in which a viewing device 860 receives one or more images in connection with a particular gaze direction of the device. In one embodiment, the device has received information from the server 840 explaining what tile streams are available and information about each tile stream such as spatial coverage and bitrate. The gaze direction of the viewing device 860 is captured and the viewing device 860 requests a subset of tile streams from the server 840. The viewing device 860 receives tile streams and composes a single bitstream from the tiles streams and feeds the bitstream to the decoder 450. The viewing device 860 has the necessary information about the tile streams and knows how it composed the single bitstream. Therefore, the viewing device 860 is able to convert the decoded pictures into pictures that are suitable for rendering.

In another embodiment, the gaze direction of the viewing device 860 is captured and sent to the server 840. The server 840 composes a single bitstream from the stored tiles streams and sends the bitstream to the decoder 450. The necessary information how to convert the decoded pictures into pictures that are suitable for rendering is sent along with the bitstream. The viewing device 860 decodes the pictures (decoder 450), converts them and renders them (rendering 870). Optionally, the viewing device 860 or client continuously sends requests for what bitrate the stream should have.

An encoder 400 and decoder 450 could be implemented using a suitable combination of hardware and/or software. Accordingly, an encoder comprising one or more processors and e.g. one or more memories is configured to carry out the methods according to the embodiments herein. Figure 30 is a schematic diagram showing components of an encoder 400. The processor 500 could be any combination of one or more of a suitable central processing unit (CPU), multiprocessor, microcontroller, digital signal processor (DSP), application specific integrated circuit (ASIC) etc., capable of executing software instructions contained in a computer program 520 stored in one or more memories 515. The one or more memories 515 is for instance a data buffer, Read-Write Memory (RWM), Read-Only memory (ROM), random-access memory (RAM), cache memory, flash memory devices, optical storage devices, etc. The encoder could comprise an interface 540 with an input and/or output circuit 530 configured to transmit a bitstream, e.g., via transmitter or transceiver 550.

As shown in Figure 31 an encoder 400 could have units, modules, or circuits configured to carry out methods described herein. For instance, the encoder 400 comprises a define partition unit 600 that defines a partition structure that divides a picture into picture segments as described herein. The encoder 400 comprises an encode unit 610 that encodes picture segments in accordance with the partition structure to generate a plurality of coded picture segments. The encoder 400 comprises a generate unit 620 that generates a bitstream comprising the plurality of coded picture segments and information indicating the partition structure used to divide the picture into the plurality of picture segments.

Further, a decoder 450 comprising one or more processors and e.g. one or more memories is configured to carry out the methods according to the embodiments herein. Figure 32 is a schematic diagram showing components of decoder 450. The processor 510 could be any combination of one or more of a suitable central processing unit (CPU), multiprocessor, microcontroller, digital signal processor (DSP), application specific integrated circuit (ASIC) etc., capable of executing software instructions contained in a computer program 550 stored in one or more memories 540. The one or more memories 540 is for instance, a data buffer, Read- Write (RWM) Memory or Read-Only (ROM) memory. The decoder 450 comprises an interface 545 with an input and/or output circuit 535 configured to receive a bitstream, e.g., via receiver or transceiver 560.

As shown in Figure 33, a decoder 450 could have units, modules, or circuits configured to carry out methods described herein. For instance, the decoder 450 comprises a receive unit 630 configured to receive a bitstream. The decoder 450 comprises a derive unit 640 configured to derive, from the information in the bitstream, a partition structure. The decoder 450 comprises a decode unit 650 configured to decode a coded picture segments received in the bitstream in accordance with the partition structure.

It should be noted that the embodiments in this disclosure comprise segments or tiles having a rectangular shape, where each segment or tile has four sides or edges. Additionally, each side or edge borders either other segments or tiles or the picture boundary. That is, the picture can be extended to a size equal to an integer number of units.

In this disclosure, a“tile” is a type of segment, and the terms“side” and“edge” can be used interchangeably.

Some of the embodiments contemplated herein are described more fully with reference to the accompanying drawings. Other embodiments, however, are contained within the scope of the subject matter disclosed herein. The disclosed subject matter should not be construed as limited to only the embodiments set forth herein; rather, these embodiments are provided by way of example to convey the scope of the subject matter to those skilled in the art.