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

Abstract

The invention relates to a loop filter method used in image code process. Wherein, it comprises: first, calculating the number of similar points in the margin of filtered code block, and according to said number and if there is frame internal block, to decide the block effective value of said code block; then according o the pixel difference between two sides of block margin, judging if the block margin is the real edge of picture; at last, processing filter treatment on the code block according to the block effective value, when it is false edge. The invention improves the flat area filtration, to obtain better effect on the block removal effect of flat area.

Description

Loop circuit filtering method during image encoding is handled

Technical field

The present invention relates to the image encoding processing technology field, relate in particular to the loop circuit filtering method in a kind of image encoding processing.

Background technology

Rapid popularization along with Development of Multimedia Technology and multimedia application, H.264 the video image compression standard of uniting the JVT of the standard formulation working group formulation that forms by MPEG and two normal structures of ITU, with its advanced technical characterstic and compression performance preferably, obtained extensive use in fields such as video conferencing, video telephone, Streaming Media, HD DVD, video monitoring, Digital Television, 3G.

H.264 compression is handled framework as shown in Figure 1, basic handling unit is 16 * 16 macro blocks, adopted advanced technologies such as multiframe reference, infra-frame prediction, many macro block modes, 4 * 4 integer transforms and quantification, loop filtering, the prediction of 1/4 picture element movement, CAVLC (based on contextual variable-length encoding algorithm) and CABAC (based on contextual arithmetic coding algorithm) entropy coding, its compression efficiency than MPEG-2, H.263, MPEG-4ASP doubles many.

In addition; China digital audio/video encoding and decoding technique standard operation group AVS organized to set up the standard series that a cover is similar to mpeg standard since 2002; comprise parts such as video encoding standard, audio coding standard, system standard, digital copyright protecting DRM standard; and issued AVSpart1 system standard, AVS part2 video standard in December, 2003, issued AVS part7 video standard in December, 2004.

H.264, in the video encoding standard such as AVS, image is all needed to do piecemeal handles, the appearance blocking effect that each piece all can be in various degree in conversion and after quantizing.The appearance of blocking effect mainly be since when quantizing piecemeal handle the quantization error that causes and cause that described blocking effect has two kinds of different forms of expression with the difference of picture material in the piecemeal scope, be mainly trapezoidal noise and lattice shape noise.This noise like will make the image border obviously even boxed area occurs.

Described trapezoidal noise appears at the strong edge of image.Because many high frequency coefficients of DCT (discrete cosine transform) are quantified as zero, make that strong edge can not be by fully digitalization in transform domain, again because the processing of the piecemeal of image, feasible continuity of passing the strong edge of block boundary can not be guaranteed, thereby the zigzag noise appears in strong edge at image, making visually has factitious data block edge, is referred to as trapezoidal noise.

Described lattice shape noise appears at the flat site of image.Flat site at image, brightness has under the situation that increases progressively or successively decrease, owing to quantize to round up when rounding, may cause DC (direct current) coefficient of transform domain to cross the decision threshold of adjacent quantized level, cause that jump in brightness appears in adjacent two pieces in reconstructed image, make the sheet profile visually to occur, be referred to as lattice shape noise.

For fear of blocking effect, intact in image encoding, store before the reference frame as back one frame, need do the processing of deblocking effect to image.In the MPEG series standard, generally adopted post-processing filter (post-filtering) to overcome blocking effect.And in H.264, then adopted the mode of loop filter (loop-filter) to overcome blocking effect.Adopt loop filter and adopt post-processing filter to compare, the former need not the special buffering area present frame that prestores, so the hardware realization is more prone to; Moreover loop filter is positioned at prediction loop, and directly therefore the impact prediction frame, can play the effect that reduces residual error coefficient.

Below will be at AVS part2[1] in the implementation of the loop filtering that uses describe, specifically comprise:

At first, computing block effect value and pixel value difference, and judged:

According to macroblock coding information calculations blocking effect value Bs, if macro block is the intraframe coding type, Bs=2 then, if macro block is the interframe encode type, and motion vector and reference frame are variant at the block boundary place, Bs=1 then, otherwise Bs=0;

Create the threshold value table of two neighbors to difference according to quantization parameter QP, the difference size that neighbor is right compares with threshold value table, judges image edge information;

Secondly, just can carry out corresponding Filtering Processing:

If the blocking effect value on a limit of current Filtering Processing is 2, and the right difference of the pixel at block boundary place then makes mean filter with several to the filtering window corresponding pixel value less than the value that defines in the threshold value table, produces two pairs of new pixel values;

If the blocking effect value on a limit of current Filtering Processing is 1, and the right difference of the pixel at block boundary place is less than the value that defines in the threshold value table, then original pixel value increased or deducts a difference, produces two pairs of new pixel values;

The if block effect value is 0, and perhaps the difference that the pixel at block boundary place is right is not then done filtering greater than the value that defines in the threshold value table.

As can be seen, ignored in the easier fact of being discovered by human eye of flat site blocking effect in the above-mentioned existing loop filtering process, thereby do not carried out the smothing filtering of hard intensity at flat site.And in said method, filter operator has only 3 grades (0,1,2), because filter operator is too single, thereby is difficult to reach good filtering effect, and then the occasion that makes corresponding method adapt to is limited.

Summary of the invention

In view of above-mentioned existing in prior technology problem, the purpose of this invention is to provide the loop circuit filtering method in a kind of image encoding processing, thereby can realize good deblocking effect, and can adapt to more deblocking effect occasion at flat site.

The objective of the invention is to be achieved through the following technical solutions:

The invention provides the loop circuit filtering method in a kind of image encoding processing, comprising:

A, statistics are treated the quantity of the encoding block borderline region similitude of filtering, and determine the blocking effect value of this encoding block according to the quantity of similitude;

B, according to the pixel difference situation on block boundary both sides, whether the decision block edge is the true edge of image;

C, for the true edge of image, then do not need filtering, and, then adopt corresponding filtering mode to carry out Filtering Processing to encoding block according to described blocking effect value for the non-true edge of image.

Described encoding block border is 8 * 8 block boundaries, or 4 * 4 block boundaries.

Described steps A comprises:

The quantity of statistical coding block boundary zone similitude, when described quantitative value during less than first predetermined value, determine that the blocking effect value is 0, when described quantitative value greater than first predetermined value during less than second predetermined value, determine that the blocking effect value is 1, when described quantitative value during, determine that the blocking effect value is 2 greater than second predetermined value.

Described steps A also comprises:

When described encoding block is Intra-coded blocks, then the blocking effect value of determining is added 1.

Described similitude is meant that pixel value difference is less than two points of predetermined value in the borderline region.

The quantity of described similitude be encoding block edge homonymy and both sides similitude to number and.

Described method also comprises:

D, setting are 0 comprising the blocking effect value, and need not to carry out Filtering Processing when the blocking effect value are 0 greater than 3 concrete blocking effect values and corresponding filter processing method thereof.

Described step D comprises:

When the blocking effect value on a limit of encoding block was 3, corresponding Filtering Processing mode produced 3 pairs of new pixel values for the filter window of correspondence pixel value being made mean filter;

When the blocking effect value on a limit of encoding block was 2, corresponding Filtering Processing mode produced 2 pairs of new pixel values for the filter window of correspondence pixel value being made mean filter;

When the blocking effect value on a limit of encoding block was 1, corresponding Filtering Processing mode produced 2 pairs of new pixel values for original pixel value being increased or deducting a difference.

Among the present invention, when the blocking effect value is 3, adopt eight tap filters to carry out Filtering Processing.

Described step B comprises:

Whether judge the right difference of the pixel of boundary less than predetermined threshold value, if, then determine it is not the true edge of image, otherwise, be defined as the true edge of image.

Described predetermined threshold value is to obtain according to look-up method, concrete threshold value is stored in two threshold value tables respectively, suppose it is respectively α, β table, the α table record thresholding of block boundary both sides pixel value differences, the β table record thresholding of inner two pixel value differences of piece, write down two one-dimension array respectively in two tables, its length all is the span of quantized value, and lookup table index adds a side-play amount for quantization step QP.

Described threshold value table is:

Index	α	β	Index	α	β	Index	α	β	Index	α	β
													0	0	0	16	4	2	32	22	6	48	46	15
1	0	0	17	4	2	33	24	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	21	7	3	37	33	8	53	54	20
												6	1	1	22	8	4	38	33	8	54	55	21
7	1	1	23	9	4	39	35	9	55	56	22
												8	1	1	24	10	4	40	35	9	56	57	23
9	1	1	25	11	4	41	36	10	57	58	23
												10	1	1	26	12	5	42	37	10	58	59	24
11	2	1	27	13	5	43	37	11	59	60	24
												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

；

Perhaps, represent α, β table, be respectively in the mode of array:

The 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

}。

Among the present invention, when the blocking effect value was 3, the described Filtering Processing of step C comprised:

Adopted (1,2,2,2,1) weight coefficient right at R0 and L0;

Adopted (1,1,2) and (1,1,4,2) weight coefficient right at R1 and L1;

Adopted (1,1,4,2) weight coefficient right at R2 and L2;

Wherein, R0, R1, R2, L0, L1, L2 are respectively the pixel value in block edge right side and left side.

As seen from the above technical solution provided by the invention, the loop filtering implementation method that the present invention adopts can be so that signal to noise ratio improves about 0.08db, and code check descends about 1.5%.Simultaneously, because having carried out improving at flat site filtering, loop filtering algorithm provided by the invention designs, so the deblocking effect of flat site is had good effect.

Also filtering progression is expanded among the present invention 4 grades (0,1,2,3), thereby also make and more refinement of filter operator be adapted to different deblocking effect occasions more.

The present invention is for texture region, owing to checked true edge before filtering, so can not produce the phenomenon that true edge is level and smooth.

Except that above-mentioned advantage, the present invention can also realize the parallel processing of macro-block level, and promptly can realizing in same macro block walks abreast simultaneously carries out Filtering Processing to many limits.

Description of drawings

Fig. 1 handles the framework schematic diagram for H.264 compressing;

Fig. 2 is a method flow diagram of the present invention;

Fig. 3 is the block boundary schematic diagram;

Fig. 4 is the edge pixel value schematic diagram of the required use of 8 tap filters on the vertical boundary.

Embodiment

The object of the present invention is to provide a kind of loop circuit filtering method, it can effectively remove blocking artifact, makes the subjective quality of decoding back image improve, and objectively Y-PSNR improves, and code check descends.

Describe below in conjunction with the specific implementation of accompanying drawing to method provided by the invention, wherein the encoding block border is 8 * 8 block boundaries, and for 4 * 4 block boundary situations, method is similar.As shown in Figure 2, specifically comprise following processing procedure:

At first, need computing block effect value strength;

Step 21: the number n um of statistics block borderline region similitude;

So-called similitude, be the point of pixel value difference less than predetermined value, described predetermined value can be set points such as 2,3, the quantity of described similitude be encoding block edge homonymy and both sides similitude to number and, as shown in Figure 3, Ln, Hn, Rn, Kn (0＜=n＜=7) are 32 points that are positioned at block boundary, and then num is calculated by following formula:

$\mathrm{num}=\underset{n=0}{\overset{7}{\mathrm{\&Sigma;}}}((\mathrm{abs}(L_n-H_n)<2)?1:0)+\underset{n=0}{\overset{7}{\mathrm{\&Sigma;}}}((\mathrm{abs}(H_n-R_n)<2)?1:0)$

(1)；

$+\underset{n=0}{\overset{7}{\mathrm{\&Sigma;}}}((\mathrm{abs}(R_n-K_n)<2)?1:0)$

Wherein, abs (A-B) is the absolute value operation of getting A-B.For horizontal boundary, take same way, to three groups on horizontal direction points do difference again summation obtain num, symbol "? " represent that its preceding face amount greater than 0 o'clock, gets the A value among the A:B, otherwise get the B value.

Step 22: determine to treat to be specially the blocking effect value of filter block according to the number of similitude of statistics:

1, as if num＜4, strength=0;

2, as if num＞=4 and num＜16, strength=1;

3, as if num＞=16, strength=2;

Certainly, also can determine corresponding blocking effect value according to number and other threshold values of similitude;

Step 23: judge whether to exist piece in the frame, if then execution in step 24, otherwise, direct execution in step 25:

The piece that is specially both sides, if block edge has at least one to be to adopt intraframe coding method, then is judged as to have piece in the frame.

Step 24: if a piece is arranged is piece in the frame, then corresponding strength is added 1, and afterwards, execution in step 25 again.

Secondly, just need to carry out Filtering Processing according to above-mentioned definite blocking effect value.

Because filtering is smooth operation, thus true edge might be made smoothing processing, in order effectively to prevent this operation, should be before filtering at doing true and false differentiation when leading edge.

Step 25: judge and work as whether leading edge is true edge that if then execution in step 26, otherwise process finishes;

Be specially the pixel value difference on both sides, computing block border, whether judge this difference greater than a certain preset threshold, if greater than, then be the image real border, do not need to carry out filtering, otherwise, be defined as carrying out the border of Filtering Processing;

For judging whether to be true edge, at first need to calculate described pixel value difference, specifically can adopt predefined one-dimensional linear operator.Its operand is predefined one dimension window.For example, set a pair of point on block boundary both sides, R0 as shown in Figure 3 and L0, and the inner two couples of some L0 of fritter and L1, R0 and R1.

In addition, also need to preestablish corresponding judgement threshold value, concrete threshold value is stored in two threshold value tables respectively, suppose it is respectively α, β table, the α table record thresholding of difference of block boundary both sides pixels, the β table record thresholding of difference of inner two points of piece, write down two one-dimension array respectively in two tables, its length all is the span of quantized value, is 64 by default.

And the concrete value of threshold value is to determine according to the quantization step QP that square quantizes in described two tables.Simultaneously, adjust the dynamics that filtering is carried out for convenience, can also artificially set two side-play amounts of adjusting filtering.At this moment, the array indexing value of described threshold value table is exactly that quantized value adds side-play amount, and under the default conditions, the value of these two side-play amounts all is 0.When looking into this table, then with QP and side-play amount and obtain two values, a constraint is done in the back, obtains two index value indexA and indexB.At α, find out corresponding α in the β table according to these two values, the β value is as the threshold value of differentiating;

Wherein α, β table can be as follows with array form and tabular form respectively:

(1) the 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	α	β	Index	α	β	Index	α	β	Index	α	β
												0	0	0	16	4	2	32	22	6	48	46	15
1	0	0	17	4	2	33	24	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	21	7	3	37	33	8	53	54	20
												6	1	1	22	8	4	38	33	8	54	55	21
7	1	1	23	9	4	39	35	9	55	56	22
												8	1	1	24	10	4	40	35	9	56	57	23
9	1	1	25	11	4	41	36	10	57	58	23
												10	1	1	26	12	5	42	37	10	58	59	24
11	2	1	27	13	5	43	37	11	59	60	24
												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 Fig. 4, illustrate: subtract each other with R0 and L0, obtain a difference C1 with the example on the horizontal direction; Subtract R1 with R0, obtain a difference C2; Subtract L1 with L0, obtain a difference C3; Afterwards, if C1 less than α, C2 less than β, thinks then that the edge feature of image is not in the edge of piece less than β and C3, can carry out filtering; Otherwise then do not advance filtering.

When determining to carry out Filtering Processing, then the blocking effect value of determining according to step 22 is carried out corresponding Filtering Processing, specifically can adopt following processing procedure to realize:

Step 26: judge whether described blocking effect value is 0, if then process finishes, this is because need not to carry out Filtering Processing when the blocking effect value is 0, otherwise, execution in step 27;

Step 27: judge whether described blocking effect value is 1, if then execution in step 210, otherwise, continue execution in step 28:

Step 28: judge whether described blocking effect value is 2, if then execution in step 29, otherwise, continue execution in step 211;

Step 29: employing blocking effect value is that 1 o'clock Filtering Processing mode is carried out Filtering Processing, promptly adopts the difference filtering mode to carry out Filtering Processing.

Step 210: employing blocking effect value is that 2 o'clock Filtering Processing mode is carried out Filtering Processing, promptly adopts the mean filter mode to carry out Filtering Processing.

Step 211: employing blocking effect value is that 3 o'clock Filtering Processing mode is carried out Filtering Processing, promptly adopts the mean filter mode to carry out Filtering Processing.

The mean filter mode that adopts among the present invention is with to two pairs of pixels of block boundary (4 points altogether) assignment again, uses the adjustment of window filtering formula realization to pixel value, and its corresponding output valve is determined jointly by 8 points among Fig. 3; Described difference filtering also is that four points are adjusted, and it adds original pixel value or deduct a difference, thereby the gap of two pixel values is dwindled, and makes that visually blocking effect is eliminated.

To the concrete Filtering Processing process of each blocking effect value correspondence be illustrated below.

The filter operator one that adopts among the present invention has 3 grades, respectively corresponding 3 different nonzero block effect value strength.Owing to the piece that is 8 * 8, so filter window can expand to 8 tap filters.As shown in Figure 4, corresponding concrete operator is described below:

(1) when strength=1, adopt the difference filtering mode to carry out filtering:

Described difference DELTA 1=IClip (C0, C0, ((R0-L0) * 3+ (L1-R1)+4)＞＞3);

Then after the filtering, L0=L0+ Δ 1, R0=R0-Δ 1;

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, adopt the mean filter mode to carry out 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, and " d=a? b:c " implication for as a d=b greater than 0 time, otherwise d=c.

(3) when strength=3, still adopt the mean filter mode to carry out 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).

C0 reduces parameter list CLIP_TAB table by filtering of inquiry to obtain in the computing formula of above Filtering Processing, is index with QP at first, and the value that checks in CLIP_TAB is C0, and described CLIP_TAB table can be expressed as with array form and form:

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	21	1	37	2	53	6
								6	0	22	1	38	3	54	7
7	0	23	1	39	3	55	7
								8	0	24	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

Be to be the explanation that example is carried out during the front is described with Filtering Processing process to vertical boundary, for horizontal boundary, its filter window is in vertical state, but corresponding filter operator is consistent with the filter operator on the vertical boundary, corresponding Filtering Processing process is also identical, so no longer describe in detail.

In sum, the loop filtering implementation method that the present invention adopts can improve on the objective performance about 0.08db, and code check descends about 1.5%.

And, because having carried out improving at flat site filtering, loop filtering algorithm provided by the invention designs, so the deblocking effect of flat site is had good effect.

Filtering progression is expanded to 4 grades (0,1,2,3), make more refinement of filter operator to be adapted to different deblocking effect occasions more.

For texture region, owing to before filtering, checked true edge, so can not produce the phenomenon that true edge is level and smooth.

Can realize the parallel processing of macro-block level, i.e. the Filtering Processing on many limits of parallel processing simultaneously in same macro block.

The above; only for the preferable embodiment of the present invention, but protection scope of the present invention is not limited thereto, and anyly is familiar with those skilled in the art in the technical scope that the present invention discloses; the variation that can expect easily or replacement all should be encompassed within protection scope of the present invention.Therefore, protection scope of the present invention should be as the criterion with the protection range of claim.

Claims (13)

1, the loop circuit filtering method during a kind of image encoding is handled is characterized in that, comprising:

A, statistics are treated the quantity of the encoding block borderline region similitude of filtering, and determine the blocking effect value of this encoding block according to the quantity of similitude;

B, according to the pixel difference situation on block boundary both sides, whether the decision block edge is the true edge of image;

C, for the true edge of image, then do not need filtering, and, then adopt corresponding filtering mode to carry out Filtering Processing to encoding block according to described blocking effect value for the non-true edge of image.

2, the loop circuit filtering method during image encoding according to claim 1 is handled is characterized in that described encoding block border is 8 * 8 block boundaries, or 4 * 4 block boundaries.

3, the loop circuit filtering method during image encoding according to claim 1 is handled is characterized in that described steps A comprises:

The quantity of statistical coding block boundary zone similitude, when described quantitative value during less than first predetermined value, determine that the blocking effect value is 0, when described quantitative value greater than first predetermined value during less than second predetermined value, determine that the blocking effect value is 1, when described quantitative value during, determine that the blocking effect value is 2 greater than second predetermined value.

4, the loop circuit filtering method during image encoding according to claim 3 is handled is characterized in that described steps A also comprises:

When described encoding block is Intra-coded blocks, then the blocking effect value of determining is added 1.

5, according to the loop circuit filtering method in claim 1,2, the 3 or 4 described image encodings processing, it is characterized in that described similitude is meant that pixel value difference is less than two points of predetermined value in the borderline region.

6, the loop circuit filtering method in handling according to claim 1,2,3 or 4 described image encodings is characterized in that, the quantity of described similitude be encoding block edge homonymy and both sides similitude to number and.

7, according to the loop circuit filtering method in claim 1,2, the 3 or 4 described image encodings processing, it is characterized in that described method also comprises:

D, setting are 0 comprising the blocking effect value, and need not to carry out Filtering Processing when the blocking effect value are 0 greater than 3 concrete blocking effect values and corresponding filter processing method thereof.

8, the loop circuit filtering method during image encoding according to claim 7 is handled is characterized in that described step D comprises:

When the blocking effect value on a limit of encoding block was 3, corresponding Filtering Processing mode produced 3 pairs of new pixel values for the filter window of correspondence pixel value being made mean filter;

When the blocking effect value on a limit of encoding block was 2, corresponding Filtering Processing mode produced 2 pairs of new pixel values for the filter window of correspondence pixel value being made mean filter;

When the blocking effect value on a limit of encoding block was 1, corresponding Filtering Processing mode produced 2 pairs of new pixel values for original pixel value being increased or deducting a difference.

9, the loop circuit filtering method during image encoding according to claim 8 is handled is characterized in that, when the blocking effect value is 3, adopts eight tap filters to carry out Filtering Processing.

10, according to the loop circuit filtering method in claim 1,2, the 3 or 4 described image encodings processing, it is characterized in that described step B comprises:

Whether judge the right difference of the pixel of boundary less than predetermined threshold value, if, then determine it is not the true edge of image, otherwise, be defined as the true edge of image.

11, the loop circuit filtering method during image encoding according to claim 10 is handled, it is characterized in that, described predetermined threshold value is to obtain according to look-up method, concrete threshold value is stored in two threshold value tables respectively, suppose it is respectively α, β table, the α table record thresholding of block boundary both sides pixel value differences, the β table record thresholding of inner two pixel value differences of piece, two one-dimension array have been write down respectively in two tables, its length all is the span of quantized value, and lookup table index adds a side-play amount for quantization step QP.

12, the loop circuit filtering method during image encoding according to claim 10 is handled is characterized in that described threshold value table is: Index α β Index α β Index α β Index α β 0 0 0 16 4 2 32 22 6 48 46 15 1 0 0 17 4 2 33 24 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 21 7 3 37 33 8 53 54 20 6 1 1 22 8 4 38 33 8 54 55 21 7 1 1 23 9 4 39 35 9 55 56 22 8 1 1 24 10 4 40 35 9 56 57 23 9 1 1 25 11 4 41 36 10 57 58 23 10 1 1 26 12 5 42 37 10 58 59 24 11 2 1 27 13 5 43 37 11 59 60 24 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 ；

Perhaps, represent α, β table, be respectively in the mode of array:

The 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

}。

13, the loop circuit filtering method during image encoding according to claim 8 is handled is characterized in that when the blocking effect value was 3, the described Filtering Processing of step C comprised:

Adopted (1,2,2,2,1) weight coefficient right at R0 and L0;

Adopted (1,1,2) and (1,1,4,2) weight coefficient right at R1 and L1;

Adopted (1,1,4,2) weight coefficient right at R2 and L2;

Wherein, R0, R1, R2, L0, L1, L2 are respectively the pixel value in block edge right side and left side.

Images