A known technique for pixel colour value encoding is run length coding. The technique allows sets of adjacent pixel values to be coded more compactly by specifying the colour once and then the count of the number (n) of identical pixels instead of repeating the colour code (n) times.

Further information about run length coding may be found in, for example, "Principles of Interactive Computer Graphics"by W M Newman and R F Sproul, International Student Edition, 1979, pp 287-289, pub McGraw-Hill, ISBN 0-07-066455-2.

An example of an encoding and decoding system making use of run length coding to reduce the necessary volume of data to code whole or partial image frames is given in International patent application WO 96/25010 commonly assigned with the present application. In the example, a method is described of encoding pixel colour values for a digital video image frame in which each of up to 15 different colours within the image is assigned a colour value. A predominant colour (i. e. the most commonly occurring pixel colour value) is identified for the image frame and, in a first embodiment, each pixel having one of the 14 colours other than the predominant colour is separately coded simply identifying its respective colour value (the codes 0010 to 1111 being used), with runs of three or more successive pixels of the predominant colour being run-length encoded.

A further code (0000 0011 followed by a four bit colour code), similar in arrangement to that indicating a run, is provided to allow a change in the

specified predominant colour during the course of a frame. In a further embodiment, runs of all colours are run-length encoded but with a shorter coding scheme for runs of the predominant colour or, in a still further embodiment, a small range of predominant colours.

A principle use for these coding schemes is to improve efficiency of coding for certain classes of image material, in particular subtitling or other text boxes to be displayed overlaid on a video image. For such applications, the particular aim is to achieve at least a reasonable degree of compression without incurring large overheads due to complexity. For the particular case of recorded video, where the subtitling or other data may be stored as, for example, a separate file on an optical disc, the requirement for compression increases due to the limitations of available storage space whilst efficiency must not be degraded to the point where decoding and regeneration of the data becomes a significant factor affecting playback performance.

It is accordingly an object of the present invention to provide a coding methodology for pixel colour values which provides good compression without causing undue delays in decoding.

It is a further object to provide a means for encoding and decoding according to such a methodology.

In accordance with the present invention there is provided a method of encoding pixel colour values for a digital video image frame in which each different colour within the image is assigned a colour value, wherein a predominant colour is identified for the image frame, and runs of at least three successive pixels of the predominant colour are encoded as a first code indicating a run and a second code indicating the run length; characterised in that runs of successive pixels having a colour other than the predominant colour are encoded as successive iterations of a code containing only the respective colour value when the run length is less than or equal to a threshold value and as a first code indicating a run, a second code indicating a run length, and a third code identifying the colour value

when the threshold value is exceeded.

As will become apparent hereinafter, the present invention is optimised for limited colour schemes and especially for the coding of text blocks, although it is not limited to such. In an embodiment to be described, four different colours including the predominant colour may be supported with each pixel having a colour other than the predominant colour being separately coded as a 2-bit code: the predominant colour is likely but not certain to be a background colour and the three remaining colours may be respectively assigned to foreground and to two intermediate levels between foreground and background for the purposes of anti-aliasing. By keeping the non-predominant colour codes short, the repetitive iterations for short to medium length runs (i. e. below the threshold-which threshold may differ for different colours) do not generate an excessive bit load.

For increased savings, one of the colours other than the predominant colour may be selected, with a range of run lengths between minimum and maximum values being specified for that colour, wherein: runs of the selected colour below the minimum value are coded as separate iterations of the colour code; runs of the selected colour between the minimum and maximum values are coded as a first code indicating a run of that colour and a second code indicating the length; and runs of the selected colour above the maximum value are coded as a first code indicating a run, a second code indicating a run length, and a third code identifying the colour value of the selected colour. With this shortened code for short to medium runs (e. g. 4 to 11 pixels) in what may be the text colour for subtitling applications for example, the present applicants have found that further bit-savings of up to 5% in relation to the Digital Video Broadcast 2-bit coding standard may be achieved.

In a practical arrangement, all codes for a pixel or a run preferably comprise an integer number of bit-pairs for ease of decoder implementation.

A further code may be provided, wherein subsequent to the placing of the further code in a stream of pixel colour codes, the colour specified for the

immediately preceding pixel is applied to all further pixels to the end of a display line. This further code is suitably only used if run-length specification of the remainder of the line or repeated iterations of the colour code per pixel do not require fewer bits to specify, that is to say if the further code is the most economical option.

A still further code may be provided in a stream of pixel colour codes, with the code identifying to a host decoder a predetermined and stored pattern of pixel colour values to be called up and applied to the following pixels.

Also in accordance with the present invention there is provided video image encoding apparatus arranged to encode pixel colour values for a digital video image frame by assigning to each different colour within the image a respective colour value, wherein a predominant colour is identified for the image frame; characterised in that the apparatus includes means arranged to identify runs of successive pixels having a colour other than the predominant colour and encode the same as successive iterations of a code containing only the respective colour value when the run length is less than or equal to a threshold value and as a first code indicating a run, a second code indicating a run length, and a third code identifying the colour value when the threshold value is exceeded.

Further in accordance with the present invention there is provided a video image signal comprising encoded frames of pixel colour values, a storage medium carrying such a video image signal, and video image playback apparatus as defined in the attached claims, the disclosure of which is incorporated herein and to which reference should now be made.

Preferred embodiments of the present invention will now be described, by way of example only, and with reference to the accompanying drawings in which: Figure 1 shows a three colour screen icon on a background of a fourth (predominant) colour;

Figures 2 to 5 represent respective proportions of the number of per image pixel colour runs for differing run lengths and for each of the four colours of Figure 1; Figure 6 is a table of colour and run codes allocated according to an embodiment of the present methodology; and Figure 7 is a block schematic diagram of part of a receiver device for decoding for display screen messages encoded using the scheme of Figure 6.

Beginning with Figure 1, this schematically represents a text character formed by pixels of a given colour on a background of pixels of a second colour. Around the character, two concentric bands of pixels are shown: whilst these bands may be in contrasting colours to the foreground and/or the background, it is preferred for simple subtitling and captioning purposes that they be differently proportioned mixes of foreground and background (for example 70% background to 30% foreground for one of them with the proportions reversed for the other) to enable at least crude anti-aliasing at boundaries between blocks of foreground and background pixels. As will be well understood, this is of particular value where the edge of characters do not coincide with pixel boundaries which could result in "staircase"effects on sloping edges if some degree of foreground/background mixing is not used.

In the following description, this colour arrangement will be concentrated upon, with the background identified as colour (C)"0", the two anti-aliasing colours identified as respectively colour"1"and colour"2", and the foreground colour of the character as colour"3".

The basis for the selection of the particular codes (to be described below with reference to the table of Figure 6) is to take account of the likelihood of pixel colour runs in terms of likely length and frequency for each of the four colours in dependence on the subject matter encoded. As previously mentioned, a particular concern is for subtitling or captioning

boxes where the box will generally comprise a rectangle of the background colour within which text messages printed in the foreground colour appear.

Figures 2 to 5 show the result of tests over subtitling texts featuring Chinese and Latin characters of various font sizes with distribution of run- lengths of the four colours 0,1,2, and 3 indicated in respective ones of the Figures.

For each of these Figures, the vertical axis represents the percentage of the total number of pixel runs in all colours for the image or coded segment, and the horizontal axis represents run length in pixels, up to a maximum length of 284 pixels. This horizontal axis is subdivided into three consecutive ranges for run length, with short runs from 1 to 3 pixels in length, medium runs from 4 to 11 pixels length, and long runs from 12 to 284 pixels. As shown by Figure 2, there is a fairly constant distribution for background pixel runs of from 1 to around 10 or 11 pixels (short and medium runs), each representing around 2% of the total runs. For long runs, those of up to around 30 to 35 pixels each represent less than 1 % of the total, with the remainder of the long range (up to 284) covering less than 0.01 %.

For the anti-aliasing colours 1 and 2 of Figures 3 and 4 respectively, the pattern is substantially the same for both, with single pixels representing about 20% of the total runs, dropping to around 1.5% for runs of two pixels, to below 0.5% for medium runs, to below 0.05% for runs in the range 12 to around 35 pixels, and substantially zero for any runs greater than this length. For the foreground colour 3, as shown in Figure 5 the pattern begins with a peak at around 22% in the short run range: the level of this peak remains largely constant although (as shown) its location along the run length axis may vary in dependence on factors such as the font size.

After this initial peak, the pattern drops to around 2% for mid-range runs and long runs dropping to zero as for the anti-aliasing colours of Figures 3 and 4.

In known schemes such as that of the proposed digital video

broadcast (DVB) standard, it is assumed that the runs in the medium range have equal probabilities of occurrence for all colours. The present applicants have recognised that the background and the foreground colours occur far more often than the anti-aliasing ones and in the present scheme they are treated as"preferred"colours. Thus they are given separate codewords of smaller size to encode their run-length only and not their colour value (as is done in DVB 2-bit schemes).

Another characteristic of the present scheme is that short runs of the colours that predominate in the short runs, that is to say colours 1,2, and 3, are coded as successive iterations of their colour value, whereas the background colour 0 which is predominant in all other regions has separate codewords for each case.

The detailed encoding scheme is as shown in the table of Figure 6 and provides short (2-bit) codes for the non-predominant colours: as these are more likely to appear as single or repeated iterations, it produces a saving to give these a shorter code than the four bits used to code a single pixel of the predominant colour and six bits used to encode a pair of these pixels. As previously mentioned, the background and foreground colours 0, 3, are preferred and respective codes are provided to code runs of each in the medium range, with common further codes distinguished only by the addition of the pixel colour code to specify runs of any pixel in the long range. These codewords that can provide long runs of any of the colours permit encoding at minimum cost for possible occurrences of border lines around the text.

As shown in the penultimate row of Figure 6, the scheme provides a codeword to denote an end-of-line if other codewords cannot accommodate all the pixels until the end of scan line efficiently enough. In other words, where the same pixel colour is specified to the end of a scanline, the code "00 00 00 00"will be used if this takes less space than a sequence of colour code iterations or the specification of a run termination at the last scanline pixel.

As an option, one extra codeword (the last row of Figure 6) is available to encode, if needed, one predetermined (and stored) pattern of colours. Such a pattern is preferably long to provide savings from specifying it as a block and preferably occurs relatively frequently in a particular whole or partial picture frame for which it is selected.

A particular feature of this scheme is that it can be efficiently read by the decoder because all codes are multiples of 2-bits in length and each decision in a decoding tree can be made from reading 2 bits at a time: systems that try to read an odd number of bits at a time suffer because of difficulties of extraction from the storage medium. This happens because the decoder must make a more laborious bit selection out of each codeword.

Methods that extract 2,4 or 8 bits at a time have been found by us to be more efficient.

A block schematic diagram of a part of a decoder apparatus for handling overlay graphics (OG) coded according to the present scheme is shown in Figure 7. The input to the apparatus is an MPEG2 program stream which is passed to a demultiplexer 20. Following a selection process in the demultiplexer, a stream of subtitle data packets are output to a transport buffer 22 with timing data being passed straight through and into a further buffer 24. From the transport buffer 22, the undecoded data is passed to a decoder stage 26 where it is decoded and separated into basic overlay graphics codes (as in Figure 6) which are supplied to a display buffer 28, and detailed specification of the colours cross-referenced by means of a colour look-up table (CLUT) to the 2-bit colour codes, which specification is downloaded to CLUT buffer 30. The final stage is a display encoder 32 which takes the data from the display buffer 28 and, by reference to the CLUT in buffer 30, generates to subtitle display images to be mixed with other images such as in the form of an overlay to a full-motion video sequence.

From reading the present disclosure, other modifications will be apparent to persons skilled in the art. Such modifications may involve other

features which are already known in the field of video signal encoding systems and devices and component parts thereof and which may be used instead of or in addition to features already described herein. Although claims have been formulated in this application to particular combinations of features, it should be understood that the scope of the disclosure of the present invention also includes any novel feature or any novel combination of features disclosed herein either explicitly or implicitly or any generalisation thereof, whether or not it relates to the same invention as presently claimed in any claim and whether or not it mitigates any or all of the same technical problems as does the present invention. The applicants hereby give notice that new claims may be formulated to such features and/or combinations of features during the prosecution of the present application or of any further application derived therefrom.