CN1867075A - Loop filtering method in image coding processing - Google Patents

Abstract

The invention relates to a loop filter method in image coding processing. The present invention mainly includes: first, counting the number of similar points in the boundary area of the coding block to be filtered, and determining the blockiness value of the coding block according to the number of similar points and whether there is an intra-frame block; then, according to the pixels on both sides of the block boundary In the case of difference value, judge whether the edge of the block is the real edge of the image; finally, for the real edge of the image, no filtering is required, and for the non-real edge of the image, the corresponding filtering method is used to filter the coding block according to the block effect value . Since the loop filtering algorithm provided by the present invention is designed to improve the filtering of the flat area, it has a good effect on the deblocking effect of the flat area.

Description

Loop Filtering Method in Image Coding Processing

technical field

The present invention relates to the technical field of image coding processing, in particular to a loop filtering method in image coding processing.

Background technique

With the development of multimedia technology and the rapid promotion of multimedia applications, the H.264 video image compression standard formulated by JVT, a standard formulation working group jointly formed by MPEG and ITU, is characterized by its advanced technical characteristics and better compression It has been widely used in conference TV, videophone, streaming media, HD DVD, video surveillance, digital TV, 3G and other fields.

The compression processing framework of H.264 is shown in Figure 1. The basic processing unit is 16×16 macroblock, which adopts multi-frame reference, intra-frame prediction, multi-macroblock mode, 4×4 integer transformation and quantization, loop filtering, 1/4 pixel motion prediction, CAVLC (context-based variable length coding algorithm) and CABAC (context-based arithmetic coding algorithm) entropy coding and other advanced technologies, its compression efficiency is higher than MPEG-2, H.263, MPEG-4ASP more than double.

In addition, AVS, China's digital audio and video codec technology standard working group, has organized and formulated a series of standards similar to MPEG standards since 2002, including video coding standards, audio coding standards, system standards, digital copyright protection DRM standards, etc. Part, and released the AVS part1 system standard and AVS part2 video standard in December 2003, and released the AVS part7 video standard in December 2004.

In video coding standards such as H.264 and AVS, images need to be divided into blocks, and each block will have block effects to varying degrees after transformation and quantization. The occurrence of block effect is mainly caused by the quantization error caused by block processing during quantization. The block effect has two different manifestations depending on the image content within the block range, mainly trapezoidal noise and lattice noise. This type of noise will make the edges of the image noticeable or even blocky areas.

The described trapezoidal noise occurs at the strong edges of the image. Since many high-frequency coefficients of DCT (discrete cosine transform) are quantized to zero, strong edges cannot be fully digitized in the transform domain, and due to the block processing of images, the continuity of strong edges across block boundaries cannot be guaranteed. , so that jagged noise appears at the strong edge of the image, which makes the visually unnatural data block edge, which is called trapezoidal noise.

Said lattice noise occurs in flat areas of the image. In the flat area of the image, when the brightness is increasing or decreasing, due to the rounding of the quantization rounding, the DC (direct current) coefficient in the transform domain may exceed the decision threshold of the adjacent quantization level, resulting in the reconstruction of the image. There is a sudden change in brightness between two adjacent blocks, which makes a flaky outline appear visually, which is called lattice noise.

In order to avoid block effects, after the image is encoded and stored as a reference frame for the next frame, it is necessary to remove the block effect of the image. In the series of MPEG standards, post-filtering is commonly used to overcome blocking effects. In H.264, a loop-filter method is used to overcome block effects. Compared with the post-processing filter, the loop filter does not need a special buffer to pre-store the current frame, so the hardware implementation is easier; moreover, the loop filter is located in the prediction loop and can directly affect the prediction frame, so , can play a role in reducing the residual coefficient.

The implementation of loop filtering used in AVS part2 [1] will be described below, including:

First, calculate the block effect value and the pixel difference value, and make a judgment:

Calculate the block effect value Bs according to the coding information of the macroblock, if the macroblock is an intra-frame coding type, then Bs=2, if the macroblock is an inter-frame coding type, and there is a difference between the motion vector and the reference frame at the block boundary, then Bs= 1, otherwise Bs=0;

Create a threshold table for the difference between two adjacent pixel pairs according to the quantization coefficient QP, compare the difference between the adjacent pixel pairs with the threshold table, and judge the image edge information;

Secondly, the corresponding filtering processing can be performed:

If the block effect value of one side of the current filtering process is 2, and the difference between the pixel pairs at the block boundary is less than the value defined in the threshold table, then the pixel values corresponding to the filter window are used for mean filtering to generate two pairs new pixel value;

If the block effect value of one side of the current filtering process is 1, and the difference between the pixel pair at the block boundary is less than the value defined in the threshold table, add or subtract a difference value to the original pixel value to generate two pairs of new pixels value;

If the blockiness value is 0, or the difference between pixel pairs at the block boundary is greater than the value defined in the threshold table, no filtering is performed.

It can be seen that the above existing loop filtering process ignores the fact that the block effect is easier to be perceived by human eyes in flat areas, and therefore does not perform high-intensity smoothing filtering for flat areas. Moreover, in the above method, the filtering operator has only three levels (0, 1, 2). Since the filtering operator is too single, it is difficult to achieve a good filtering effect, which limits the applicable occasions of the corresponding method.

Contents of the invention

In view of the problems existing in the above-mentioned prior art, the object of the present invention is to provide a loop filtering method in image coding processing, so that good deblocking effect can be achieved in flat areas, and can be adapted to more deblocking effect occasions .

The purpose of the present invention is achieved through the following technical solutions:

The present invention provides a loop filtering method in image coding processing, comprising:

A. Count the number of similar points in the boundary area of the coding block to be filtered, and determine the blockiness value of the coding block according to the number of similar points;

B. According to the pixel difference situation on both sides of the block boundary, it is judged whether the block edge is the real edge of the image;

C. For the real edge of the image, no filtering is required, and for the non-real edge of the image, the corresponding filtering method is used to perform filtering processing on the coding block according to the block effect value.

The coding block boundary is an 8×8 block boundary, or a 4×4 block boundary.

Described step A comprises:

Statistically count the number of similar points in the boundary area of the coding block. When the number value is less than the first predetermined value, determine the blockiness value as 0, and when the number value is greater than the first predetermined value and less than the second predetermined value, determine the blockiness value The value is 1, and when the quantity value is greater than the second predetermined value, the blockiness value is determined to be 2.

Described step A also includes:

When the coding block is an intra-frame coding block, add 1 to the determined blockiness value.

The similar points refer to two points in the boundary area whose pixel difference is smaller than a predetermined value.

The number of similar points is the sum of the number of pairs of similar points on the same side and both sides of the coding block edge.

The method also includes:

D. Set more than 3 specific blocking effect values and corresponding filtering processing methods, including the blocking effect value being 0, and no filtering processing is required when the blocking effect value is 0.

Described step D comprises:

When the blockiness value of one side of the coding block is 3, the corresponding filtering processing method is to use the corresponding filtering window to perform mean filtering on the pixel values to generate 3 pairs of new pixel values;

When the blockiness value of one side of the coding block is 2, the corresponding filtering processing method is to use the corresponding filtering window to perform mean filtering on the pixel values to generate 2 pairs of new pixel values;

When the blockiness value of one edge of the coding block is 1, the corresponding filtering process is to add or subtract a difference value to the original pixel value to generate 2 pairs of new pixel values.

In the present invention, when the block effect value is 3, an eight-tap filter is used for filtering processing.

Described step B comprises:

Judging whether the difference between the pixel pairs at the boundary is smaller than a predetermined threshold value, if yes, it is determined that it is not the real edge of the image, otherwise, it is determined as the real edge of the image.

The predetermined threshold value is obtained according to the table look-up method, and the specific threshold value is stored in two threshold value tables respectively, assuming that they are α and β tables respectively, and the α table records the threshold of the pixel difference on both sides of the block boundary , the β table records the threshold of the difference between two pixels inside the block. Two one-dimensional arrays are recorded in the two tables, and their lengths are the range of quantization values. The index of the table lookup is the quantization step size QP plus a bias displacement.

The threshold value table is: index alpha beta index alpha beta index alpha beta index alpha beta 0 0 0 16 4 2 32 twenty two 6 48 46 15 1 0 0 17 4 2 33 twenty four 7 49 48 16 2 0 0 18 5 3 34 26 7 50 50 17 3 0 0 19 5 3 35 28 7 51 52 18 4 0 0 20 6 3 36 30 8 52 53 19 5 0 0 twenty one 7 3 37 33 8 53 54 20 6 1 1 twenty two 8 4 38 33 8 54 55 twenty one 7 1 1 twenty three 9 4 39 35 9 55 56 twenty two 8 1 1 twenty four 10 4 40 35 9 56 57 twenty three 9 1 1 25 11 4 41 36 10 57 58 twenty three 10 1 1 26 12 5 42 37 10 58 59 twenty four 11 2 1 27 13 5 43 37 11 59 60 twenty four 12 2 1 28 15 5 44 39 11 60 61 25 13 2 2 29 16 5 45 39 12 61 62 25 14 3 2 30 18 6 46 42 13 62 63 26 15 3 2 31 20 6 47 44 14 63 64 27 ;

Or, express the α and β tables in the form of arrays, respectively:

α table is:

byte ALPHA_TABLE[64]＝

{

0,0,0,0,0,0,1,1,

1, 1, 1, 2, 2, 2, 3, 3,

4, 4, 5, 5, 6, 7, 8, 9,

10, 11, 12, 13, 15, 16, 18, 20,

22, 24, 26, 28, 30, 33, 33, 35,

35, 36, 37, 37, 39, 39, 42, 44,

46, 48, 50, 52, 53, 54, 55, 56,

57, 58, 59, 60, 61, 62, 63, 64

};

(2) The β table is:

byte BETA_TABLE[64]＝

{

0,0,0,0,0,0,1,1,

1, 1, 1, 1, 1, 2, 2, 2,

2, 2, 3, 3, 3, 3, 4, 4,

4, 4, 5, 5, 5, 5, 6, 6,

6, 7, 7, 7, 8, 8, 8, 9,

9, 10, 10, 11, 11, 12, 13, 14,

15, 16, 17, 18, 19, 20, 21, 22,

23, 23, 24, 24, 25, 25, 26, 27

}.

In the present invention, when the blocking effect value is 3, the filtering process described in step C includes:

(1, 2, 2, 2, 1) weight coefficient pairs are adopted for R0 and L0;

(1, 1, 2) and (1, 1, 4, 2) weight coefficient pairs are used for R1 and L1;

(1, 1, 4, 2) weight coefficient pairs are adopted for R2 and L2;

Among them, R0, R1, R2, L0, L1, L2 are the pixel values on the right side and left side of the block edge respectively.

It can be seen from the above-mentioned technical solution provided by the present invention that the implementation method of loop filtering adopted in the present invention can increase the signal-to-noise ratio by about 0.08db and reduce the code rate by about 1.5%. At the same time, since the loop filtering algorithm provided by the present invention is designed to improve the flat area filtering, it has a good effect on the deblocking effect of the flat area.

In the present invention, the number of filtering stages is extended to 4 stages (0, 1, 2, 3), thus making the filtering operator more refined and more suitable for different occasions of deblocking effect.

For the texture area, the present invention does not smooth the real edge because the real edge is checked before filtering.

In addition to the above-mentioned advantages, the present invention can also realize parallel processing at the macroblock level, that is, it can simultaneously perform filtering processing on multiple edges in parallel within the same macroblock.

Description of drawings

Fig. 1 is a schematic diagram of the H.264 compression processing framework;

Fig. 2 is the method flowchart of the present invention;

Figure 3 is a schematic diagram of block boundaries;

Fig. 4 is a schematic diagram of boundary pixel values required by an 8-tap filter on a vertical boundary.

Detailed ways

The purpose of the present invention is to provide a loop filtering method, which can effectively remove the block effect, so that the subjective quality of the decoded image is improved, objectively, the peak signal-to-noise ratio is improved, and the code rate is reduced.

The specific implementation of the method provided by the present invention will be described below with reference to the accompanying drawings, wherein the encoding block boundary is an 8×8 block boundary, and the method is similar for the case of a 4×4 block boundary. As shown in Figure 2, it specifically includes the following processing procedures:

First, the block effect value strength needs to be calculated;

Step 21: Count the number num of similar points in the block boundary area;

The so-called similar points are points whose pixel value difference is less than a predetermined value, and the predetermined value can be set values such as 2, 3, etc., and the number of the similar points is a pair of similar points on the same side and both sides of the edge of the coding block The sum of the numbers, as shown in Figure 3, Ln, Hn, Rn, Kn (0<=n<=7) are 32 points positioned at the block boundary, then num is calculated by the following formula:

$\mathrm{<span\; class="notranslate">\; num</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; L</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ?</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; :</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; h</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ?</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; :</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>$

(1);

$<span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 7</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; -</span>{\mathrm{<span\; class="notranslate">\; K</span>}}_{\mathrm{<span\; class="notranslate">\; no</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; <</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; ?</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; :</span>\mathrm{<span\; class="notranslate">\; 0</span>}<span\; class="notranslate">\; )</span>$

Among them, abs(A-B) is the operation of taking the absolute value of A-B. For the horizontal boundary, take the same method, take the difference of the three groups of points in the horizontal direction and then sum to get num, the symbol "?" means that when the previous value is greater than 0, take the A value in A:B, otherwise take the B value .

Step 22: Determine the block effect value of the block to be filtered according to the number of statistical similar points, specifically:

1. If num<4, strength=0;

2. If num>=4 and num<16, strength=1;

3. If num>=16, strength=2;

Of course, the corresponding block effect value can also be determined according to the number of similar points and other threshold values;

Step 23: Determine whether there is an intra-frame block, if yes, then perform step 24, otherwise, directly perform step 25:

Specifically, if at least one of the blocks on both sides of the edge of the block adopts an intra-frame encoding method, it is determined that there is an intra-frame block.

Step 24: If a block is an intra-frame block, add 1 to the corresponding strength, and then perform step 25.

Secondly, it is necessary to perform filtering processing according to the above determined blockiness value.

Since filtering is a smoothing operation, it is possible to smooth the real edge. In order to effectively prevent this operation, the true and false judgment should be made for the current edge before filtering.

Step 25: judge whether the current edge is a real edge, if yes, execute step 26, otherwise, the process ends;

Specifically, calculate the pixel difference on both sides of the block boundary, and judge whether the difference is greater than a certain set threshold. If it is greater, it is the real boundary of the image and does not need to be filtered. Otherwise, it is determined as the boundary that needs to be filtered;

In order to judge whether it is a real edge, it is first necessary to calculate the pixel difference value, specifically, a preset one-dimensional linear operator can be used. Its operation object is a preset one-dimensional window. For example, set a pair of points on both sides of the block boundary, such as R0 and L0 shown in Figure 3, and two pairs of points L0 and L1, R0 and R1 inside the small block.

In addition, it is also necessary to pre-set the corresponding discrimination threshold value. The specific threshold value is stored in two threshold value tables, assuming that they are α and β tables respectively. The α table records the threshold of the difference between the pixels on both sides of the block boundary. , the β table records the threshold of the difference between two points inside the block. Two one-dimensional arrays are recorded in the two tables, and their lengths are the range of quantized values, which is 64 in the default state.

Moreover, the specific value of the threshold value in the two tables is determined according to the quantization step size QP of block quantization. At the same time, in order to facilitate the adjustment of the strength of the filtering execution, two offsets for adjusting the filtering can also be artificially set. At this time, the array index value of the threshold value table is the quantization value plus the offset. In a default state, the values of the two offsets are both 0. When looking up the table, use the sum of QP and offset to get two values, and then make a constraint to get two index values indexA and indexB. According to these two values, find out the corresponding α, β value in the α, β table as the threshold value of discrimination;

Among them, the α and β tables can be in the form of an array and a list as follows:

(1) α table is:

byte ALPHA_TABLE[64]＝

{

0,0,0,0,0,0,1,1,

1, 1, 1, 2, 2, 2, 3, 3,

4, 4, 5, 5, 6, 7, 8, 9,

10, 11, 12, 13, 15, 16, 18, 20,

22, 24, 26, 28, 30, 33, 33, 35,

35, 36, 37, 37, 39, 39, 42, 44,

46, 48, 50, 52, 53, 54, 55, 56,

57, 58, 59, 60, 61, 62, 63, 64

};

(2) The β table is:

byte BETA_TABLE[64]＝

{

0,0,0,0,0,0,1,1,

1, 1, 1, 1, 1, 2, 2, 2,

2, 2, 3, 3, 3, 3, 4, 4,

4, 4, 5, 5, 5, 5, 6, 6,

6, 7, 7, 7, 8, 8, 8, 9,

9, 10, 10, 11, 11, 12, 13, 14,

15, 16, 17, 18, 19, 20, 21, 22,

23, 23, 24, 24, 25, 25, 26, 27

}. index alpha beta index alpha beta index alpha beta index alpha beta 0 0 0 16 4 2 32 twenty two 6 48 46 15 1 0 0 17 4 2 33 twenty four 7 49 48 16 2 0 0 18 5 3 34 26 7 50 50 17 3 0 0 19 5 3 35 28 7 51 52 18 4 0 0 20 6 3 36 30 8 52 53 19 5 0 0 twenty one 7 3 37 33 8 53 54 20 6 1 1 twenty two 8 4 38 33 8 54 55 twenty one 7 1 1 twenty three 9 4 39 35 9 55 56 twenty two 8 1 1 twenty four 10 4 40 35 9 56 57 twenty three 9 1 1 25 11 4 41 36 10 57 58 twenty three 10 1 1 26 12 5 42 37 10 58 59 twenty four 11 2 1 27 13 5 43 37 11 59 60 twenty four 12 2 1 28 15 5 44 39 11 60 61 25 13 2 2 29 16 5 45 39 12 61 62 25 14 3 2 30 18 6 46 42 13 62 63 26 15 3 2 31 20 6 47 44 14 63 64 27

As shown in Figure 4, take an example in the horizontal direction: Subtract R0 and L0 to get a difference C1; subtract R1 from R0 to get a difference C2; subtract L1 from L0 to get a difference C3; after that, if If C1 is less than α, C2 is less than β, and C3 is less than β, it is considered that the feature edge of the image is not at the edge of the block, and filtering can be performed; otherwise, no filtering is performed.

When it is determined that filtering processing is required, corresponding filtering processing is performed according to the block effect value determined in step 22, specifically, the following processing can be used to achieve:

Step 26: judging whether the blockiness value is 0, if yes, the process ends, this is because no filtering process is required when the blockiness value is 0, otherwise, perform step 27;

Step 27: Determine whether the blockiness value is 1, if yes, execute step 210, otherwise, continue to execute step 28:

Step 28: judge whether the blockiness value is 2, if yes, execute step 29, otherwise, continue to execute step 211;

Step 29: Perform filtering processing using a filtering processing method when the block effect value is 1, that is, perform filtering processing using a difference filtering method.

Step 210: Perform filtering processing using the filtering processing method when the block effect value is 2, that is, performing filtering processing using the mean value filtering method.

Step 211: Perform filtering processing using the filtering processing method when the block effect value is 3, that is, performing filtering processing using the mean value filtering method.

The average value filtering method adopted in the present invention is to reassign two pairs of pixels (4 points in total) on the block boundary, and use the window filter formula to adjust the pixel value, and its corresponding output value is jointly determined by 8 points in Fig. 3 ; The difference filter also adjusts four points, which adds or subtracts a difference to the original pixel value, thereby narrowing the gap between the two pixel values, so that the visual block effect is eliminated.

The specific filtering process corresponding to each blocking effect value will be illustrated below with an example.

The filter operator used in the present invention has three levels in total, corresponding to three different non-zero block effect values strength respectively. Since it is an 8×8 block, the filtering window can be extended to an 8-tap filter. As shown in Figure 4, the corresponding specific operators are described as follows:

(1) When strength=1, use the difference filtering method to filter:

The difference Δ1=IClip(-C0, C0, ((R0-L0)*3+(L1-R1)+4)>>3);

After filtering, L0=L0+Δ1, R0=R0-Δ1;

Δ2=IClip(-C0, C0, ((L0-L1)*3+(L2-R0)+4)>>3)

L1=L1+Δ2

Δ3=IClip(-C0, C0, ((R1-R0)*3+(L0-R2)+4)>>3)

R1=R1-Δ3.

(2) When strength=2, use the mean value filtering method for filtering:

R0=aq? (R1+R0+L0+R0+2)>>2: ((R1<<1)+R0+L0+2)>>2;

L0=ap? (L0+L1+R0+L0+2)>>2: ((L1<<1)+L0+R0+2)>>2;

R1=aq? (R2+R0+L0+R1+2)>>2: R1;

L1=ap? (L2+L1+L0+R0+2)>>2: L1;

Wherein, "d=a?b:c" means that when a is greater than 0, d=b, otherwise d=c.

(3) When strength=3, still use the mean filtering method for filtering:

R0=aq? ((R1<<1)+R2+((R0+L0)<<1)+L1+4)>>3: (R1+R0+L0+R0+2)>>2;

L0=ap? (L2+(L1<<1)+((R0+L0)<<1)+R1+4)>>3: (L0+L1+R0+L0+2)>>2;

R1=aq? (R0+R1+L0+R1+2)>>2: R1;

L1=ap? (L1+L1+L0+R0+2)>>2: L1;

R2=aq? (R0+R1+(R2<<2)+(R3<<1)+4)>>3: R2;

L2=ap? (L0+L1+(L2<<2)+(L3<<1)+4)>>3: L2;

Wherein, aq=(abs(R0-R2)<β); ap=(abs(L0-L2)<β).

In the calculation formula of the above filtering process, C0 is obtained by querying a filtering and trimming parameter table CLIP_TAB table. First, QP is used as an index, and the value checked in CLIP_TAB is C0. The CLIP_TAB table can be used in array form and table form. Expressed as:

byte CLIP_TAB[64]＝

{

0,0,0,0,0,0,0,0,

0,0,0,0,0,0,0,0,

1, 1, 1, 1, 1, 1, 1, 1,

1, 1, 1, 1, 1, 1, 2, 2,

2, 2, 2, 2, 2, 2, 3, 3,

3, 3, 3, 3, 3, 4, 4, 4,

5, 5, 5, 6, 6, 6, 7, 7,

7, 7, 8, 8, 8, 9, 9, 9

}. index CI index CI index CI index CI 0 0 16 1 32 2 48 5 1 0 17 1 33 2 49 5 2 0 18 1 34 2 50 5 3 0 19 1 35 2 51 6 4 0 20 1 36 2 52 6 5 0 twenty one 1 37 2 53 6 6 0 twenty two 1 38 3 54 7 7 0 twenty three 1 39 3 55 7 8 0 twenty four 1 40 3 56 7 9 0 25 1 41 3 57 7 10 0 26 1 42 3 58 8 11 0 27 1 43 3 59 8 12 0 28 1 44 3 60 8 13 0 29 1 45 4 61 9 14 0 30 2 46 4 62 9 15 0 31 2 47 4 63 9

In the previous description, the filtering process of the vertical boundary is taken as an example. For the horizontal boundary, the filter window is in the vertical state, but the corresponding filter operator is consistent with the filter operator on the vertical boundary, and the corresponding filter processing The process is also the same, so it will not be described in detail.

To sum up, the implementation method of loop filtering adopted in the present invention can improve the objective performance by about 0.08db, and the code rate can be reduced by about 1.5%.

Moreover, since the loop filtering algorithm provided by the present invention is designed to improve the flat area filtering, it has a good effect on the deblocking effect of the flat area.

Extending the number of filtering stages to 4 stages (0, 1, 2, 3) makes the filtering operator more refined and more suitable for different occasions of deblocking effect.

For the texture area, since the real edge is checked before filtering, the phenomenon of smoothing the real edge will not occur.

Parallel processing at the macroblock level can be realized, that is, the filtering processing of multiple edges can be processed in parallel in the same macroblock at the same time.

The above is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed in the present invention can easily think of changes or Replacement should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be determined by the protection scope of the claims.

Claims (13)

1. A loop filtering method in image coding processing, characterized in that, comprising: A. Count the number of similar points in the boundary area of the coding block to be filtered, and determine the blockiness value of the coding block according to the number of similar points; B. According to the pixel difference situation on both sides of the block boundary, it is judged whether the block edge is the real edge of the image; C. For the real edge of the image, no filtering is required, and for the non-real edge of the image, the corresponding filtering method is used to perform filtering processing on the coding block according to the block effect value. 2. The loop filtering method in image coding processing according to claim 1, wherein the coding block boundary is an 8×8 block boundary, or a 4×4 block boundary. 3. The loop filtering method in image coding processing according to claim 1, wherein said step A comprises: Statistically count the number of similar points in the boundary area of the coding block. When the number value is less than the first predetermined value, determine the blockiness value as 0, and when the number value is greater than the first predetermined value and less than the second predetermined value, determine the blockiness value The value is 1, and when the quantity value is greater than the second predetermined value, the blockiness value is determined to be 2. 4. The loop filtering method in image coding processing according to claim 3, wherein said step A further comprises: When the coding block is an intra-frame coding block, add 1 to the determined blockiness value. 5. The loop filtering method in image coding process according to claim 1, 2, 3 or 4, characterized in that, said similar points refer to two points in the boundary area whose pixel difference value is less than a predetermined value. 6. The loop filtering method in image coding processing according to claim 1, 2, 3 or 4, characterized in that the number of similar points is the number of similar point pairs on the same side and both sides of the edge of the coding block of and. 7. The loop filtering method in image coding processing according to claim 1, 2, 3 or 4, characterized in that said method further comprises: D. Set more than 3 specific blocking effect values and corresponding filtering processing methods, including the blocking effect value being 0, and no filtering processing is required when the blocking effect value is 0. 8. The loop filtering method in image coding processing according to claim 7, wherein said step D comprises: When the blockiness value of one side of the coding block is 3, the corresponding filtering processing method is to use the corresponding filtering window to perform mean filtering on the pixel values to generate 3 pairs of new pixel values; When the blockiness value of one side of the coding block is 2, the corresponding filtering processing method is to use the corresponding filtering window to perform mean filtering on the pixel values to generate 2 pairs of new pixel values; When the blockiness value of one edge of the coding block is 1, the corresponding filtering process is to add or subtract a difference value to the original pixel value to generate 2 pairs of new pixel values. 9. The loop filtering method in image coding processing according to claim 8, characterized in that, when the blockiness value is 3, an eight-tap filter is used for filtering processing. 10. The loop filtering method in image coding process according to claim 1, 2, 3 or 4, characterized in that said step B comprises: Judging whether the difference between the pixel pairs at the boundary is smaller than a predetermined threshold value, if yes, it is determined that it is not the real edge of the image, otherwise, it is determined as the real edge of the image. 11. The loop filtering method in image coding processing according to claim 10, characterized in that, the predetermined threshold value is obtained according to a table look-up method, and the specific threshold value is stored in two threshold values respectively. The value table is assumed to be α and β tables respectively. The α table records the threshold of the pixel difference on both sides of the block boundary, and the β table records the threshold of the difference between two pixels inside the block. Two one-dimensional arrays are recorded in the two tables respectively. Its length is the value range of the quantization value, and the look-up index is the quantization step size QP plus an offset. 12. The loop filtering method in image coding processing according to claim 10, characterized in that, the threshold value table is: index alpha beta index alpha beta index alpha beta index alpha beta 0 0 0 16 4 2 32 twenty two 6 48 46 15 1 0 0 17 4 2 33 twenty four 7 49 48 16 2 0 0 18 5 3 34 26 7 50 50 17 3 0 0 19 5 3 35 28 7 51 52 18 4 0 0 20 6 3 36 30 8 52 53 19 5 0 0 twenty one 7 3 37 33 8 53 54 20 6 1 1 twenty two 8 4 38 33 8 54 55 twenty one 7 1 1 twenty three 9 4 39 35 9 55 56 twenty two 8 1 1 twenty four 10 4 40 35 9 56 57 twenty three 9 1 1 25 11 4 41 36 10 57 58 twenty three 10 1 1 26 12 5 42 37 10 58 59 twenty four 11 2 1 27 13 5 43 37 11 59 60 twenty four 12 2 1 28 15 5 44 39 11 60 61 25 13 2 2 29 16 5 45 39 12 61 62 25 14 3 2 30 18 6 46 42 13 62 63 26 15 3 2 31 20 6 47 44 14 63 64 27 ; Or, express the α and β tables in the form of arrays, respectively: Alpha table is: byte ALPHA_TABLE[64]={ 0,0,0,0,0,0,1,1, 1, 1, 1, 2, 2, 2, 3, 3, 4, 4, 5, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 18, 20, 22, 24, 26, 28, 30, 33, 33, 35, 35, 36, 37, 37, 39, 39, 42, 44, 46, 48, 50, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64}; (2) The β table is: byte BETA_TABLE[64]={ 0,0,0,0,0,0,1,1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 6, 6, 6, 7, 7, 7, 8, 8, 8, 9, 9, 10, 10, 11, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 23, 24, 24, 25, 25, 26, 27 }. 13. The loop filtering method in image coding processing according to claim 8, wherein when the blockiness value is 3, the filtering processing in step C includes: (1, 2, 2, 2, 1) weight coefficient pairs are adopted for R0 and L0; (1, 1, 2) and (1, 1, 4, 2) weight coefficient pairs are used for R1 and L1; (1, 1, 4, 2) weight coefficient pairs are adopted for R2 and L2; Among them, R0, R1, R2, L0, L1, L2 are the pixel values on the right side and left side of the block edge respectively.

Images