DESCRIPTION

METHOD, APPARATUS AND INTEGRATED CIRCUIT CAPABLE OF REDUCING IMAGE RINGING NOISE

Techn ical F ie I d

The invention relates to a method, an apparatus and an integrated circuit capable of reducing ringing noise, more particularly to a method, an apparatus and an integrated circuit capable of reducing image ringing noise.

Background Art

Nowadays, while high compression of image data can be achieved by image data encoding techniques, the pre-processing and post-processing of image data appear to be ever more important. Many techniques have been proposed for filtering noise generated when image data is acquired and encoded.

As shown in Figure 1, U.S. Patent No. 5,818,964 discloses a method and an apparatus for selecting an adaptive filter for image data. This prior art has been adopted by the Motion Picture Experts Group-4 (MPEG-4) standard to eliminate associated ringing noise. The noise filtering process 1 of this prior art consists of two sub- processes, that is, an index value generating unit 2 and a selective local smoothing unit 3. The index value generating unit 2 has a threshold value determining unit 4 and a binary index value unit 5. The selective local smoothing unit 3 has a filter selecting unit 6 and an adaptive filtering unit 7. First, an input image is inputted into the index value generating unit 2. Subsequently, the threshold value

determining unit 4 calculates the maximum and minimum gray levels for every block consisting of 8 x 8 pixels in the input image to determine a threshold value. In the binary index value unit 5, the threshold value is compared with each pixel value in a corresponding block so as to generate binary indices. In the filter selecting unit 6, the binary indices are used as a basis to select a smoothing filter. The adaptive filtering unit 7 is for performing a smoothing filter process so as to adjust the gray level of a block.

However, since the abovementioned prior art uses the maximum and minimum pixel values in a corresponding block to determine the threshold value, textures and edges that only present local variations cannot be effectively detected. Therefore, the textures and the edges are likely to be classified as noise and smoothed, thereby reducing the sharpness in images. In addition, real noise may possibly be classified as textures or edges and thus not be removed. Therefore, a solution is necessary.

D i s c l o s u re of I n v e n t i o n

Therefore, the object of the present invention is to provide a method capable of reducing image ringing noise so as to improve quality of images or videos.

Accordingly, the method capable of reducing image ringing noise of the present invention comprises the following steps. First, a line spread function (LSF) value of each pixel's luminance in an input image is calculated. Then, the input image is divided into a plurality of blocks, and an average value and a maximum value of the line spread function values for all pixels in each block are calculated. Subsequently, based on the average value and the

maximum value of the line spread function values of each block, it is determined whether a block is an interfered flat block or a non- interfered block. Next, the pixels in the interfered flat block are set as noisy pixels. Then, it is determined whether each pixel in the non- interfered block is an edge pixel or a non-edge pixel, and non-edge pixels in the non-interfered block are set as noisy pixels. Subsequently, smoothing the noisy pixels thus set is performed, and smoothed pixels are combined with the edge pixels for output as an output image. Another object of the present invention is to provide an apparatus capable of reducing image ringing noise so as to improve quality of images or videos.

Accordingly, the apparatus capable of reducing image ringing noise of the present invention comprises a line spread function calculating unit, a characteristic value calculating unit, a block category determining unit, a noisy pixel setting unit, an edge pixel determining unit, and a pixel smoothing unit. The line spread function calculating unit is for calculating a line spread function value of each pixel's luminance in an input image. The characteristic value calculating unit is for dividing the input image into a plurality of blocks, and for calculating an average value and a maximum value of the line spread function values for all pixels in each block. The block category determining unit is for determining whether a block is an interfered flat block or a non-interfered block based on the average value and the maximum value of the line spread function values of each block. The noisy pixel setting unit is for setting the pixels in the interfered flat block as noisy pixels. The edge pixel determining unit is for determining whether each pixel in

the non-interfered block is an edge pixel or a non-edge pixel, and for setting non-edge pixels in the non-interfered block as noisy pixels. The pixel smoothing unit is for smoothing noisy pixels thus set, and for combining smoothed pixels with the edge pixels for output as an output image.

Yet another object of the present invention is to provide an integrated circuit capable of reducing image ringing noise so as to improve quality of images or videos.

Accordingly, the integrated circuit capable of reducing image ringing noise of the present invention comprises a line spread function calculating unit, a characteristic value calculating unit, a block category determining unit, a noisy pixel setting unit, an edge pixel determining unit, and a pixel smoothing unit. The line spread function calculating unit is for calculating a line spread function value of each pixel's luminance in an input image. The characteristic value calculating unit is for dividing the input image into a plurality of blocks, and for calculating an average value and a maximum value of the line spread function values for all pixels in each block. The block category determining unit is for determining, based on the average value and the maximum value of the line spread function values of each block, whether a block is an interfered flat block or a non-interfered block. The noisy pixel setting unit is for setting the pixels in the interfered flat block as noisy pixels. The edge pixel determining unit is for determining whether each pixel in the non- interfered block is an edge pixel or a non-edge pixel, and for setting non-edge pixels in the non-interfered block as noisy pixels. The pixel smoothing unit is for smoothing noisy pixels thus set, and for combining smoothed pixels with the edge pixels for output as an

output image.

B ri ef D e s cri pt i o n of D raw i n g s

The aforesaid and other technical contents, features and advantages of the present invention will be clearly presented in the following detailed description of two preferred embodiments with reference to the accompanying drawings, of which:

Figure 1 is a function block diagram, illustrating a method and an apparatus for selecting an adaptive filter for image data in the prior art;

Figure 2 is a function block diagram, illustrating a first preferred embodiment of an apparatus capable of reducing ringing noise according to the present invention;

Figure 3 is a flow chart, illustrating a first preferred embodiment of a method capable of reducing ringing noise according to the present invention;

Figure 4 is a schematic diagram to illustrate a horizontal block category determining module in the embodiment of the present invention adding absolute values of edge pixel luminance differences in a horizontal direction of each block to serve as a corresponding change trend value Trend _x for the respective block, where the edge pixels refer to the horizontally leftmost and rightmost pixels in the block;

Figure 5 is a schematic diagram to illustrate a vertical block category determining module in the embodiment of the present invention adding absolute values of edge pixel luminance differences in a vertical direction of each block to serve as a corresponding change trend value Trend _y for the respective block, where the edge

pixels refer to the vertically topmost and bottommost pixels in the block;

Figure 6 is a function block diagram, illustrating a second preferred embodiment of an apparatus capable of reducing ringing noise according to the present invention;

Figure 7 is a flow chart, illustrating a second preferred embodiment of a method capable of reducing ringing noise according to the present invention;

Figure 8 is a schematic diagram, illustrating a vague line detection step in a horizontal direction in the second preferred embodiment of the present invention; and

Figure 9 is a schematic diagram, illustrating a vague line detection step in a vertical direction in the second preferred embodiment of the present invention.

B e s t M o d e f o r C a r ry i n g O u t t h e I n v e n t i o n

Before the present invention is described in greater detail, it should be noted herein that like elements are denoted by the same reference numbers throughout the following disclosure. Referring to Figure 2, the first preferred embodiment of an apparatus capable of reducing ringing noise according to the present invention includes a line spread function calculating unit, a characteristic value calculating unit, a macro block processing unit 83, a block category determining unit, a noisy pixel setting unit 85, an edge pixel determining unit, and a pixel smoothing unit 87.

In the first preferred embodiment, some of the components of the apparatus according to the present invention can process images in a horizontal and a vertical direction, respectively, but

should not be limited thereto. Therefore, as shown in Figure 2, the line spread function calculating unit includes a horizontal line spread function calculating module 811 and a vertical line spread function calculating module 812. The characteristic value calculating unit includes a horizontal characteristic value calculating module 821 and a vertical characteristic value calculating module 822. The block category determining unit includes a horizontal block category determining module 841 and a vertical block category determining module 842. The edge pixel determining unit includes a horizontal edge pixel determining module 861 and a vertical edge pixel determining module 862.

The line spread function calculating unit is for calculating a line spread function value of each pixel's luminance in an input image. The characteristic value calculating unit is for dividing the input image into a plurality of blocks, and for calculating an average value and a maximum value of the line spread function values for all pixels in each block, where each block is composed of, for example, 8x8 pixels. The macro block processing unit 83 is for dividing the input image into a plurality of macro blocks (MBs), each of which is composed of a plurality of the blocks, and for setting a maximum line spread function value of a macro block to be the maximum of the line spread function values of the pixels contained in the macro block, where each macro block is composed of, for example, 2x2 blocks.

The block category determining unit is for determining whether a block is an interfered flat block or a non-interfered block based on

the average value and the maximum value of the line spread function values of each block and the maximum line spread function value of the macro block containing the block.

The noisy pixel setting unit 85 is for setting the pixels in the interfered flat block as noisy pixels.

The edge pixel determining unit is for determining whether each pixel in the non-interfered block is an edge pixel or a non-edge pixel, and for setting non-edge pixels in the non-interfered block as noisy pixels. The pixel smoothing unit 87 is for smoothing the noisy pixels thus set, and for combining smoothed pixels with the edge pixels for output as an output image.

Referring to Figures 2 and 3, the first preferred embodiment of a method capable of reducing ringing noise according to the present invention comprises the following steps. First, as shown in step 91, using the horizontal line spread function calculating module 811 and the vertical line spread function calculating module 812, the line spread function values LSF _x (JJ) and LSF _y (i,j) of a pixel's luminance in a two-dimensional position (i,j) in an input image, i.e., luminance φ(i,j) , are respectively calculated in the horizontal and vertical directions as shown in the following Equations (1) and (2).

LSF, XU)=VJdJ) = ϊψ ¹ →jy φ(i ^~ lJ) (1) dx

LSF _y (i J) = V φ{ij) = ^hA = φ{ij)-φ{i j-l) (2)

Then, as shown in step 92, using the horizontal characteristic value calculating module 821 and the vertical characteristic value calculating module 822,

the input image is divided into a plurality of blocks, and average values Avg _x , Avg _y and maximum values Max _x , Max _y of the line spread function values in the horizontal and vertical directions for all pixels in each block are calculated.

Subsequently, using the macro block processing unit 83, the input image is divided into a plurality of macro blocks, and a maximum line spread function value of a macro block, MB _Max _xy , is set to be the maximum of the line spread function values of the pixels contained in the macro block. In the first preferred embodiment, since each macro block is composed of 2x2 = 4 blocks, MB_Max _xy for the macro block can be defined by the following Equation (3). MB JMax _xy - V _i=14 max(Mαx, _w ,Mαx, _ω ) ( 3 )

With further reference to Figures 4 and 5, next as shown in step 93, using the horizontal block category determining module 841 and the vertical block category determining module 842, the absolute values of edge pixels' luminance differences in every block are added horizontally and vertically to serve as corresponding change trend values Trend _x , Trend _y for the block, where the edge pixels refer to the leftmost and rightmost pixels in the horizontal direction in the block, and the topmost and bottommost pixels in the vertical direction in the block. In the first preferred embodiment, since each block is composed of 8x8 pixels, Tr end _x , Trend can be obtained using the

following Equations (4), (5), where the left top corner of each block has a coordinate (χ _o ,y _o ) on the image, and abs denotes the absolute value function.

Trend _x = ∑abs(φ(iJ)-φ(7.+lJ)) (4) j=y _B i _o +7 • Trend _y = ∑abs(φ(i,j)-φ(iJ + 7)) (5)

Subsequently, based on the following conditional determination equations (6), (7), the horizontal block category determining module 841 and the vertical block category determining module 842 determine, horizontally and vertically, whether the maximum line spread function value of the macro block- containing the block, MB_Max _xy , is greater than a macro block threshold value MB_Thr, whether the maximum values of the line spread function values of the block Max _x , Maχ _y are smaller than a flat threshold value FlatThr , and whether the change trend values Trend _x , Trend _y o f the block's luminance are smaller than a trend threshold value TrendThr . If the result of the determination is yes, then the block is an interfered flat block. Otherwise, the block is a non-interfered block.

In the horizontal direction: if (MB _Max _xy ≥MB _ Thr) and (Max _x < FlatThr) and (Trend _x < TrendThr) the block is an interfered flat block; else the block is a non- interfered block (6)

In the vertical direction: // (MB _Max _xy ≥ MB _ Thr) and (Max _y < FlatThr) and (Trend _y < TrendThr) the block is an interfered flat block; else the block is a non- interfered block (7)

When a block is determined to be an interfered flat block, as shown in step 94, the noisy pixel setting unit 85 sets the pixels in the interfered flat block as noisy pixels.

On the other hand, when a block is determined to be a non-interfered block, as shown in step 95, the horizontal edge pixel determining module 861 and the vertical edge pixel determining module 862 determine, horizontally and vertically, whether each pixel in the non-interfered block is an edge pixel or a non-edge pixel, that is, first according to the following conditional determination equations (8), (9), for each block, it is determined horizontally and vertically whether the average values Avg _x , Avg _y of the line spread function values of the block are greater than an average threshold, AvgThr , or whether the maximum values Mαχ _x , Mαx _y of the line spread function values of the block are greater than a maximum threshold value MαxThr . If the determination result is yes, edge threshold values EdgeThr _x , EdgeThr _y are respectively set to be the smaller of one half of the maximum value Mαx _x or Mαχ _y of the line spread function values and a predefined maximum edge value MαxEdge .

Otherwise, the edge threshold values EdgeThr _x , EdgeThr _y are the maximum values Max _x , Max _y of the line spread function values .

In the horizontal direction:

if (Avg _x > AvgThr) or (Max _x > MaxThr)

EdgeThr _x = mm(MaxEdge ₇ Max _x 12); else

EdgeThr _x =Max _x

In the vertical Direction: if (A vg _y > AvgThr) or (Max _y > MaxThr)

EdgeThr = mm( (MMaaxxEEddgςree, M Maaxx / 122\)_; (9) else

EdgeThr _y =Max _y

Subsequently, based on the following conditional determination equations (10), (11 ), for a pixel located at position (i,j), it is determined horizontally and vertically whether its line spread function values LSF _x (iJ) , LSF _y (iJ) are greater than the edge threshold value EdgeThr _x , EdgeThr _y respectively. If the determination result is yes, then the pixel is an edge pixel. Otherwise, the pixel is a non-edge pixel.

In the horizontal direction: if (LSF _x QJ) >EdgeThr _x ) the pixel is an edge pixel; else the pixel is a non-edge pixel In the vertical direction:

if (LSF _y (i,j)>EdgeThr _y ) the pixel is an edge pixel;

(11) else the pixel is a non-edge pixel

Furthermore, when a pixel in the non-interfered block is determined to be a non-edge pixel, as shown in step 96, the horizontal edge pixel determining module 861 and the vertical edge determining module 862 further set the non-edge pixel as a noisy pixel.

Next, as shown in step 98, using the pixel smoothing unit 87, smoothing is performed on each noisy pixel thus set in steps 94 and 96. In the first preferred embodiment, a low pass filter (LPF) as shown in the following matrix (12) can be used to perform smoothing on each noisy pixel thus set in steps 94 and 96, in which one of possible setting methods is to set a value (b) as two times a value (a), and a value (c) as two times the value (b), but is not limited thereto, as long as the value (c) is enhanced to be distinguishable from the values (a) and (b).

1

(12)

4α + 4b + c

Next, as shown in step 99, the noisy pixels that have been smoothed by the pixel smoothing unit 87 and the edge pixels can be combined together, and outputted as an output image with reduced ringing noise and maintained texture and pattern information.

As shown in Figures 6 and 7, aside from all of the components in the first preferred embodiment of the apparatus according to the present invention shown in Figure 2, the second preferred embodiment of the apparatus according to the present invention further comprises a vague line detection unit that includes a horizontal vague line detection module 881 and a vertical vague I ine detection module 882 for determining horizontally and vertically whether the noisy pixels thus set are line pixels. If the determination result is yes, smoothing on the noisy pixels by the pixel smoothing unit 87 is not necessary. Steps 91~96 for the second preferred embodiment of the method according to the present invention illustrated in Figure 7 are identical to those illustrated in Figure 6. The difference lies in that, in the second preferred embodiment, as shown in step 97, it is necessary to use the horizontal vague line detection module 881 and the vertical vague line detection module 882 to determine horizontally and vertically directions whether each noisy pixel is a line pixel. If the determination result is yes, then the noisy pixel does not require smoothing. On the other hand, if the determination result is no, then, as shown in step 98, smoothing is performed on each noisy pixel that is a non-line pixel using the pixel smoothing unit 87. In the second preferred embodiment, the low pass filter as shown in

the above matrix (12) can also be used to perform smoothing on the noisy pixels that are non-line pixels. Next, as shown in step 99, the noisy pixels that have been smoothed by the pixel smoothing unit 87, the noisy pixels that need no smoothing (i.e., the line pixels) and the edge pixels can be combined together, and outputted as an output image with reduced ringing noise and maintained texture and pattern i nf ormation . The details regarding step 97 in the second preferred embodiment are as follows. First, as shown in Equations (13), (14), horizontal and vertical line threshold values LineThr _x , LineThr _y of ^' a block are determined, where EdgeThr _x , Edgeηιr _y are horizontal and vertical edge threshold values of the block, and MaxLineThr is a predefined maximum line threshold value.

LineThr _x =Max(EdgeThr _x /2, MaxLineThr) (13)

LineThr _y = Max(EdgeThr/2, MaxLineThr) (14) With further reference to Figures 8 and 9, subsequently, determination equations (15), (16) and three 1x3 windows (its center cell being a pixel with shading) are used to check whether each noisy pixel is a line pixel. As shown in Figure 8, take the horizontal direction as an example, if the LSF _x of three consecutive windows are all greater than LineThr _x or all smaller than - LineThr _x , the noisy pixel located in the center cell is a line pixel.

In the horizontal direction: if (LSF _x (I, j) > LineThr _x and LSF _x (J + 1, j + 1) > LineThr _x and LSF _x Q - 1, / - 1) > LineThr _x )or (LSF _x (I, j) > LineThr _x and LSF _x (i + 1, j - 1) > LineThr _x and LSF _x (J - 1, j + V) > LineThrJor (LSF _x (i, j) > LineThr _x and LSF _x (i, j-Y)> LineThr _x and LSF _x (i, j + l)> LineThr _x ) or (LSF _x (i, ]) < -LineThr _x and LSF _x (i + 1, j + 1) < -LineThr _x and LSF _x (i - 1, / - 1) < -LineThr _x ) or (LSF _x (J, j) < -LineThr _x and LSF _x (i + 1, j - 1) < -LineThr _x and LSF _x (i - 1, j + 1) < -LineThr _x ) or (LSF _x (i, j) < -LineThr _x and LSF _x (i, ; - 1) < -LineThr _x and LSF _x (i, j + 1) < -LineThr _x ) the pixel is a line pixel; (15)

In the vertical direction: if (LSF _y (i, j) > LineThr _y and LSF _y (i + 1, ; + 1) > LineThr _y and LSF _y (i - 1, j - 1) > LineThr _y )or (LSF _y (i, j) > LineThr _y and LSF _y (i + 1, j - 1) > LineThr _y and LSF _y (i - 1, j + 1) > LineThr _y ) or (LSF _y (i, j) > LineThr _y and LSF _y (i - 1, j) > LineThr _y and LSF _y (i + 1, j) > LineThr _y ) or (LSF _y (i, j) < -LineThr _y and LSF _y (i + 1, j + 1) < -LineThr _y and LSF _y (i -\,]-\)< -LineThr _y ) or

(LSF _y (i, j) < -LineThr _y and LSF _y (i + 1, j - 1) < -LineThr _y and LSF _y (/ - 1 ₃ j + 1) < -LineThr _y ) or (LSF _y (i, j) < -LineThr _y and LSF _y (i - 1 ₃ j) < -LineThr _y and LSF _y (i + 1 ₃ j) < -LineThr _y ) the pixel is a line pixel;

(16)

Referring to Figures 2 and 6, furthermore, a first preferred embodiment of the integrated circuit according to the present invention comprises the horizontal line spread function calculating module 811 , the vertical line spread function calculating module 812, the horizontal characteristic value calculating module 821 , the vertical characteristic value calculating module 822, the macro block processing unit 83, the horizontal block category determining module 841 , the vertical block category determining module 842, the noisy pixel setting unit 85, the horizontal edge pixel determining module 861 , the vertical edge pixel determining module 862, and

the pixel smoothing unit 87 shown in Figure 2. Moreover, a second preferred embodiment of the integrated circuit according to the present invention comprises the horizontal line spread function calculating module 811 , the vertical line spread function calculating module 812, the horizontal characteristic value calculating module 821 , the vertical characteristic value calculating module 822, the macro block processing unit 83, the horizontal block category determining module 841, the vertical block category determining module 842, the noisy pixel setting unit 85, the horizontal edge pixel determining module 861, the vertical edge pixel determining module 862, the horizontal vague line detection module 881 , the vertical vague line detection module 882, and the pixel smoothing unit 87 shown in Figure 6.

Therefore, through the present invention, an improvement of the quality of images or videos can be achieved for the following reasons. First, through the use of line spread function values, edges and textures can be completely and accurately detected. This makes the present invention more capable of finding edges and textures with local variations compared to the aforementioned prior art. Second, in the preferred embodiments, since the images are handled separately in horizontal and vertical directions, the characteristics of horizontal and

vertical variations can be used separately without interfering with each other. Moreover, the present invention is capable of operating alone without being connected to an image decoder and merely receiving noisy images.

In sum, in the method, apparatus and integrated circuit capable of reducing image ringing noise of this invention, through the steps of calculating the line spread function values of the pixels' luminance, it can be determined whether a block is an interfered flat block or a non-interfered block, and each pixel in the interfered flat block can be set as a noisy pixel. Next, the pixels in the non-interfered block can be classified into non-edge pixels and edge pixels, and each non-edge pixel can be set as a noisy pixel. Subsequently, the noisy pixels are smoothed, and the smoothed pixels are combined with the edge pixels for output. Therefore, ringing noise is removed so that the quality of images or videos is improved. However, the abovedescribed are merely the preferred embodiments of the present invention, and should not be taken to limit the scope of implementation of the present invention. In other words, any simple equivalent variation and modification based on the claims and the contents of the specification of the present invention are included within the scope of the present invention.

Industrial Applicability

The present invention can be applied to a method, apparatus and integrated circuit capable of reducing image ringing noise.