CN103220529A - Method for achieving video coding and decoding loop filtering - Google Patents

Abstract

The present invention relates to the technical field of multimedia video coding and decoding, and in particular to a method for implementing loop filtering at a video coding and decoding end. The method includes using block division information at the coding and decoding end to obtain an image block division structure, and the loop filtering is performed at the block boundary. ; If the boundary includes the coding block, prediction block and transformation block boundary, use the mode, motion information, quantization parameters and residuals to make boundary-level filtering decisions, and finally perform pixel-level filtering. The filtering direction can be perpendicular to the boundary or according to local pixels Characteristic or block mode and boundary form other angles, the technical effect of the present invention is: this scheme can effectively improve the subjective and objective quality of video coding, on AVS2.0 embodiment reference software RD2.0-ADI, this scheme and device can improve 0.9% encoding efficiency, while reducing the number of filtering boundaries and complexity at the encoding and decoding end.

Description

A Realization Method of Video Coding and Decoding Loop Filtering

technical field

The invention relates to the technical field of multimedia video encoding and decoding, in particular to a method for loop filtering at a video encoding and decoding end.

Background technique

In the current block-based hybrid coding framework, video coding uses a rate-distortion optimization (RDO) decision for each block to select the optimal mode of the block to obtain an approximation of the best coding performance of the current coded frame. However, in video coding, because the correlation between the spatial domain and the time domain is considered, the optimal performance of the current block cannot guarantee the optimal coding performance of the entire sequence. Some coding tools such as loop filtering can improve this local optimal and bring global benefits. The influence of subjective and objective quality. By improving the block effect caused by the block division of the current frame during encoding, the subjective ringing effect and block effect are reduced. At the same time, the filtered frame will be used as a reference frame for subsequent encoding frames, thereby improving the subjective and objective encoding quality of subsequent frames. . Therefore, loop filtering technology plays a very important role in video coding.

In the existing published video coding standards, such as H.264, the loop filtering is performed on the 4x4 block boundary, and in AVS, it is performed on the 8x8 boundary. This is because in these standards, the size of the transform block is Fixed, and not larger than the size of the coding unit block (macroblock) and the prediction block, so the boundary of the transform block must include the coding unit block and the prediction block boundary. However, in the high-efficiency video coding standard and the next-generation AVS standard AVS2.0, the coding unit (Coding Unit, CU) is the basic unit of video coding. The coding unit adopts a four-fork recursive division mode, and the coding unit can reach a maximum of 64x64 (as shown in Figure 1), instead of 16x16. For each coding unit, block coding is performed by coding tools such as prediction, transform quantization, entropy coding, and post-processing. When the coding unit is making predictions, the coding unit can be divided into prediction units (Prediction Unit, PU) of different sizes, and the prediction unit is the basic unit of prediction. The division of intra-frame prediction and inter-frame prediction units in coding units is shown in Figure 2 and Figure 3 respectively. The size of prediction blocks ranges from 64x64, 64x32, 32x64, 32x32 to 8x8, and even includes asymmetric prediction blocks (AMP). At the same time, the size of the transform block is no longer fixed at 4x4 or 8x8, and is related to the size of the CU, so the size of the transform block includes from 32x32, 16x16 to 8x8. In this way, the original method of performing filtering on all 4x4 or 8x8 boundaries will no longer meet the purpose of deblocking effect of loop filtering.

Contents of the invention

Embodiments of the present invention provide a loop filtering technology, device, and system to reduce block effects, improve subjective and objective encoded video quality and encoding efficiency, and simultaneously reduce encoding and decoding complexity.

In order to achieve the above object, the present invention provides a scheme for realizing a video codec loop filter, and the specific realization steps include:

The loop filtering method comprises the steps of:

Step 1) Use the codec terminal block division information to obtain the division structure of the image block (including coding block, prediction block and transformation block), and loop filtering is performed on the boundary of the block (such as coding block, prediction block and transformation block). If the coding block If the coding residual is 0 (that is, CBP=0), then filtering is only performed at the boundary between the coding block and the prediction block; the filtering process is a predefined filtering segmentation block, and the image is divided into filtering segmentation unit blocks (such as 16x16 block size, or is the largest coding unit block or image frame, etc.), the size of the filter segmentation block can be predefined or specified in the code stream. In order to unify the filter design, the filtering order is performed on each filter segmentation block in turn, first the vertical boundary and then the horizontal boundary , that is, first filter all the vertical boundaries in the filter segment block and then filter the horizontal boundaries; at the same time, the minimum filter unit (such as 8x8 block size) is predefined, and the block boundary whose scale is smaller than the minimum filter unit skips filtering;

Step 2) If the boundary includes the coding block, prediction block and transformation block boundary, use the mode, motion information, quantization parameters and residual to make boundary-level filtering decisions,

2.1. If the decision needs to filter the boundary, proceed to the next pixel-level filtering process;

2.2. Otherwise, skip the boundary filtering; make a pixel-level filtering decision at the block boundary that needs to be filtered to determine whether each pixel row needs filtering and the filtering strength (BS);

Step 3) Finally, pixel-level filtering is performed. The filtering direction can be perpendicular to the boundary or at other angles according to the local pixel characteristics or block mode and the boundary. The block structure involved in image filtering includes but is not limited to coding blocks, prediction blocks and transformation blocks. ;Filter segmentation blocks are used to facilitate filtering and other operations, and the image is divided in a uniform size. For example, the image can be divided according to 16x16 size filter segmentation blocks.

The step 1) also includes: the boundary of the loop filter is at the boundary of the block (such as the coding block, the prediction block and the transformation block). When the coding residual of the coding block is zero (that is, CBP=0), only the coding Block and prediction block boundaries are filtered, and the minimum size of the filter is defined, such as an 8x8 block size. For example, for the AVS2 embodiment, the size of the coding block includes the size from 64x64 to 8x8. According to the RDO selection, the quadtree iterative recursive division is used, which is the basic unit of the coding process; the prediction block is further based on the prediction based on the coding block The further division of characteristics; the transformation block is also a further division based on the distribution characteristics of the residual on the basis of the coding block.

The step 1) also includes: for the entire frame of image is divided into filter segmentation unit blocks (such as 16x16 size or the largest coding unit block), the size of the filter segmentation block can be predefined or specified in the code stream, and each filter segmentation block is sequentially scanned according to the scan Filter the segmented block to filter, first check whether the boundary includes the coding block, prediction block and transformation block boundary on the vertical boundary, and on the horizontal boundary, if so, filter the boundary, otherwise skip filtering the boundary.

The step 2) also includes: according to the mode of the blocks on both sides of the filtering boundary, motion information, quantization parameters and residuals, etc., to determine whether the current block boundary needs to be filtered, if the boundary level decision needs to filter the boundary, then proceed to the next pixel level Filtering process, otherwise the boundary filtering is skipped. If the blocks P and Q on both sides of the block boundary have intra prediction mode, filtering needs to be performed; if the P and Q blocks are both inter prediction blocks of the P frame, the residuals are 0 , the reference frame is the same and each component of the motion vector is smaller than one pixel, then it is decided that the block boundary does not need to be filtered, otherwise it needs to be filtered.

The step 2) also includes: for each row of pixels on the block boundary that needs to be filtered for filtering decision, use the jump of pixels on both sides of the boundary and compare the corresponding value of the threshold table corresponding to the index obtained by the block average QP value to select the filtering strength; pixel level The value range of filtering strength is 0-4, BS=0 means no filtering is required, BS=4 means a relatively flat area, and strong smoothing filtering is required.

The step 3) also includes: selecting the corresponding FIR filter for the position of the pixel, and the filtering direction can be perpendicular to the boundary or can be at other angles according to the local pixel characteristics or block mode and the boundary; FIR filters corresponding to different filter strengths and different pixel positions The filter is used; at the same time, the texture characteristics of the local image and the mode information of the blocks on both sides of the boundary can be used to determine whether the filtering method is perpendicular to the boundary or at a certain angle to the boundary

The technical effect of the present invention is: this scheme can effectively improve the subjective and objective quality of video coding, on the reference software RD2.0-ADI of AVS2. Decoder filters boundary number and complexity.

Description of drawings

A more complete and better understanding of the invention, and many of its attendant advantages, will readily be learned by reference to the following detailed description when considered in conjunction with the accompanying drawings, but the accompanying drawings illustrated herein are intended to provide a further understanding of the invention and constitute A part of the present invention, the exemplary embodiment of the present invention and its description are used to explain the present invention, and do not constitute an improper limitation of the present invention, wherein:

FIG. 1 is a schematic diagram of the division of coding units in the prior art of AVS2;

FIG. 2 is a schematic diagram of division of transform block units in the prior art of AVS2;

FIG. 3 is a schematic diagram of AVS2 prediction block unit division in the prior art;

Figure 4 shows that the dotted line is used by AVS1, and filters all boundaries including 8x8 transform blocks, then this solution only filters on the boundaries of solid lines (coding block, prediction block and transform block);

Fig. 5 is the pixel participating in filtering at the filtering boundary;

Figure 6 shows the filtering order in the filtering segment (taking 16x16 size as an example), the vertical boundary first, and then the horizontal boundary;

Fig. 7 is a flow chart of a filtering example supported by the present invention;

Fig. 8 is a flowchart of block boundary pixel filtering in the present invention.

Detailed ways

Encoder

As shown in Figures 1 to 3, in the AVS2 embodiment of the present invention, at the encoding end, when a frame of image is predicted, transformed and quantized, entropy encoded and mode information encoded, it can be encoded after inverse quantization and inverse transformation. Reframe the frame. The mode information and the block segmentation information block of the codec end are used to obtain the division method of the current reconstructed image frame, and the minimum filter unit of the current frame is set to divide the image into filter segmentation blocks. The frame-level filtering process is to follow the process shown in Figure 7 for each filter segment block, first check all vertical boundaries in the filter segment block whether they are coding block, prediction block or transform block boundary, and if so, use mode , motion information, quantization parameters, and residuals for boundary-level filtering decisions. If the boundary-level filtering decision needs to filter the vertical boundary, the pixel-level filtering intensity decision and pixel-level filtering are performed on the vertical boundary as shown in Figure 8. The specific filtering intensity BS decision process and the filtering process under different BSs are as described above . After performing all the vertical boundary filtering in a filtering partition block, the horizontal boundary filtering is performed in a similar manner. After filtering a filter block, it is judged whether a frame has been filtered, if not, the next filter block is prepared according to the set scanning order, and the cycle is repeated until all filter blocks of the current frame are traversed.

Decoder

For the decoding end, each filter segmentation block is sequentially processed, and after prediction and compensation plus inverse quantization and inverse transformation, the residual is obtained to decode and reconstruct the block. Perform the filtering decision and filtering process similar to the encoding end to obtain the reconstructed image at the decoding end after filtering. image for display, storage, or reference to other frames.

The above-mentioned method, device and system technical solution of the embodiment of the present invention have the following beneficial effects: This patent proposes a set of loop filtering technology. When codec performs loop filtering, the division structure of image blocks (including coding blocks, prediction blocks and transformation blocks) is obtained by using the codec end block division information, and loop filtering is performed on blocks (such as coding blocks, prediction blocks and transformation blocks) ) boundary, if the coding residual of the coding block is zero (that is, CBP=0), then filtering is only performed at the boundary of the coding block and the prediction block; the filtering process is a predefined filtering segmentation block, and the image is divided into filtering segmentation unit blocks ( Such as 16x16 block size, it can also be the largest coding unit block or image frame, etc.), the size of the filter segmentation block can be predefined or specified in the code stream. In order to unify the filter design, the filtering order is sequentially for each filter segmentation block , first the vertical boundary and then the horizontal boundary, that is, first filter all the vertical boundaries in the filtering block and then filter the horizontal boundary; at the same time, the minimum filter unit is predefined, and the block boundary with a scale smaller than the minimum filter unit skips filtering; if the boundary contains Coding block, prediction block and transformation block boundaries, use the mode, motion information, quantization parameters and residuals to make boundary-level filtering decisions, if the decision needs to filter the boundary, then proceed to the next pixel-level filtering process, otherwise skip this Boundary filtering; make pixel-level filtering decisions on the threshold at the block boundary (as shown in Attached Table 1) and the local texture features of the pixels that need to be filtered to determine whether each pixel row needs filtering and the filtering strength (BS); finally use The filter (shown in Attached Table 2 or Attached Table 3) is designed to perform pixel-level filtering, and the filtering direction can be perpendicular to the boundary or at other angles according to local pixel characteristics or block patterns and the boundary, so as to better adaptively remove block effects and improve coding efficiency. This solution is based on the RD2.0 of the AVS2.0 embodiment. This solution and the device can improve the coding efficiency by 0.9%, and can reduce the number and complexity of filtering boundaries at the encoding and decoding end.

Experimental results:

The embodiment of the present invention completes technical realization on RD2.0 (the reference software of AVS2 standard). In the experiment, the general test condition of AVS2, RA configuration, is adopted. The experimental platform is Intel(R) Xeon(R) CPU X5660 2.80GHZ2.79GHZ23.9G memory. Table 2 shows the performance comparison of using the loop filtering method in the present invention compared with the loop filtering method in the anchor (AVS2RD2.0-ADI). 4kx2k, 1080p, WVGA, WQVGA, and 720p represent test sequences with resolutions of 3840x2160, 1920x1080, 832x480, 416x240, and 1280x720, respectively. The measurement of coding performance adopts BD-rate, that is, the bit rate saving in the case of the same coding quality. According to the experimental results, by adopting the above scheme of the embodiment of the present invention, an average code rate saving of 0.9% is obtained when the encoding quality is the same.

Table 2. Performance comparison (proposed filter design vs RD2-ADI)

When codec performs loop filtering, the division structure of image blocks (including coding blocks, prediction blocks and transformation blocks) is obtained by using the codec end block division information, and loop filtering is performed on blocks (such as coding blocks, prediction blocks and transformation blocks) ) boundary, if the coding residual of the coding block is 0 (that is, CBP=0), then filtering is only performed at the boundary between the coding block and the prediction block; the filtering process is a predefined filtering segmentation block, and the image is divided into filtering segmentation unit blocks ( Such as 16x16 block size, maximum coding unit block size or image frame), you can predefine or specify the size of the filter segmentation block in the code stream. In order to unify the filter design, the filtering sequence is performed on each filter segmentation block in turn, first vertically The horizontal boundary after the boundary, that is, first filter all the vertical boundaries in the filter segment block and then filter the horizontal boundary; if the boundary includes the coding block, prediction block and transformation block boundary, use the mode, motion information, quantization parameter and residual If the decision needs to filter the boundary, then proceed to the next pixel-level filtering process, otherwise skip the boundary filtering; perform pixel-level filtering on the block boundary that needs to be filtered using threshold and pixel local texture features Decision-making to determine whether each pixel row needs to be filtered and the filtering strength (BS); finally, the designed filter is used to perform pixel-level filtering. The filtering direction can be perpendicular to the boundary or can be at other angles according to local pixel characteristics or block patterns and the boundary .

When the codec performs block boundary loop filtering, the method includes that when the codec performs loop filtering after obtaining the reconstructed image, a filtering process supported by the patent of the present invention is shown in Figure 7. The steps are as follows:

1. Initialization: Use the mode information and block segmentation information blocks of the codec to obtain the block division method of the current reconstructed image frame, set the minimum filter unit of the current frame, divide the image into filter segments, and set the initial filter segment number N= 0.

2. Perform boundary filtering on each filtering block N in the image in turn according to the scanning order

2.1. First, check all the vertical boundaries of the filtering block and check whether they are coding block, prediction block or transformation block boundary. If CBP=0, check whether it is coding block or prediction block boundary; if so, use mode and motion information , quantization parameters, and residuals to make boundary-level filtering decisions.

2.2. For all vertical boundaries, if the boundary-level filtering decision needs to filter the vertical boundary, perform pixel-level filtering strength decision and pixel-level filtering for the vertical boundary.

2.3. Then, for all horizontal boundaries of the filtered segmentation block, check whether it is a coding block, a prediction block or a transformation block boundary. If CBP=0, check whether it is a coding block or a prediction block boundary; if so, use the mode and motion information , quantization parameters, and residuals to make boundary-level filtering decisions.

2.4. For all horizontal boundaries, if the boundary-level filtering decision needs to filter the horizontal boundary, perform pixel-level filtering strength decision and pixel-level filtering for the horizontal boundary.

2.5. Determine whether the filtering of the whole frame is completed, if not, the number N of the filtering block is increased, and skip to 2.1 to continue filtering of the next filtering block.

3. The whole frame is filtered

The filter segmentation block size is 16x16 (here is just an example, the specific size can be preset or explained in the code stream), the minimum filter unit is 8x8 block (as shown in Figure 6), need to use the pixel jump and threshold at the boundary The alpha beta and local pixel jump features in the table make pixel-level filter strength (BS) decisions and select FIR filters at different pixel positions under different filter strengths (as shown in Attached Table 2 or Attached Table 3). The filtering process for a specific block boundary is shown in FIG. 8 .

The filtering strength decision-making process is as follows:

If abs(p0–q0)<Alpha&&abs(p0–q0)>1, proceed to the following steps; otherwise, Bs=0.

Set the left and right flatness variables FL and FR, and both initial values are set to 0.

If abs(p0–p1)<Beta, then FL+=2;

If abs(p0–p2)<Beta, then FL++;

If abs(q0–q1)<Beta, then FR+=2;

If abs(q0–q2)<Beta, then FR++;

The sum of left and right flatness FS=FL+FR.

According to the value of FS, make the following judgments:

When FS=6, if p0=p1, and q0=q1, then the boundary strength Bs=4. Otherwise Bs=3;

When FS=5, if p0=p1, and q0=q1, then the boundary strength Bs=3. Otherwise Bs=2;

When FS=4, if FL=2, then Bs=2. Otherwise, Bs=1.

When FS=3, and abs(p1–q1)<Beta, then Bs=1. Otherwise, Bs=0.

When FS is other values, Bs=0.

Correct the boundary strength according to the encoding situation

If encoding is performed according to the frame mode, and the chrominance component is filtered, Bs is subtracted by 1:

If encoding is performed in field mode and one of the following conditions is met, Bs is decremented by 1:

Filter the luminance component in the vertical direction;

Filter the chroma components;

When BS>0, the pixel-level filtering of the pixel row is performed, and the filter is shown in the attached table 2:

Boundary filtering process when Bs is equal to 4

When the value of boundary filtering strength Bs is 4, the calculation process of p0, p1, p2, p3 and q0, q1, q2, q3 filtering is as follows (P0, P1, P2 and Q0, Q1, Q2 are the filtered values):

P0=((p0<<3)+p0+(p2<<2)+p2+(p1<<2)+(q0<<3)-q0+(q1<<2)+(q2<<1)+q2+ 16)>>5

P1=((p0<<2)+(p0<<1)+(p1<<1)+p1+(p2<<2)+q0+(q0<<1)+8)>>4

P2=(p0+(p1<<1)+p2+(p2<<1)+p3+q0+4)>>3

Q0=((q0<<3)+q0+(q2<<2)+q2+(q1<<2)+(p0<<3)-p0+(p1<<2)+(p2<<1)+p2+ 16)>>5Q1=((q0<<2)+(q0<<1)+(q1<<1)+q1+(q2<<2)+p0+(p0<<1)+8)>>4

Q2=(q0+(q1<<1)+q2+(q2<<1)+q3+p0+4)>>3

Boundary filtering process when Bs is equal to 3

When the value of boundary filtering strength Bs is 3, the calculation process of p0, p1 and q0, q1 filtering is as follows (P0, P1 and Q0, Q1 are the filtered values):

P0=(p2+(p1<<2)+(p0<<2)+(p0<<1)+(q0<<2)+q1+8)>>4

Q0=(p1+(p0<<2)+(q0<<2)+(q0<<1)+(q1<<2)+q2+8)>>4

P1=((p2<<1)+p2+(p1<<3)+(p0<<2)+q0+8)>>4

Q1=((q2<<1)+q2+(q1<<3)+(q0<<2)+p0+8)>>4

Boundary filtering process when Bs is equal to 2

When the value of boundary filtering strength Bs is 2, the calculation process of p0 and q0 filtering is as follows (P0 and Q0 are the filtered values):

P0=((p1<<1)+p1+(p0<<3)+(p0<<1)+(q0<<1)+q0+8)>>4

Q0=((p0<<1)+p0+(q0<<3)+(q0<<1)+(q1<<1)+q1+8)>>4

Boundary filtering process when Bs is equal to 1

When the value of boundary filtering strength Bs is 1, the calculation process of p0 and q0 filtering is as follows (P0 and Q0 are the filtered values):

P0=((p0<<1)+p0+q0+2)>>2

Q0=((q0<<1)+q0+p0+2)>>2

Compared with the method in the existing AVS standard, the present invention has the following innovations and beneficial effects due to the adoption of the above-mentioned technical solution:

First, the division structure of the image block (including coding block, prediction block and transformation block) is obtained by reusing the block division information of the codec end, and the loop filtering is performed on the boundary of the coding block, prediction block and transformation block. If the CBP of the coding block= 0, the filter is only performed on the boundary of the coding block and the prediction block. This can improve the coding efficiency and reduce the number of filtering boundaries, thereby reducing the complexity of the coding and decoding end. The dotted line in Figure 4 shows that the original filtering scheme of AVS1 performs filtering on all 8x8 transform block boundaries, and the solid line represents the actual coding block. , The division method of the prediction block and the transformation block, this scheme only needs to filter at the solid line, so that the number of filtering boundaries is greatly reduced, for example, for a 32x32 block, filtering at all 8x8 block boundaries, the number of filtering boundaries is 32 8 pixels Boundary, while filtering only at the block boundary only needs to filter at 8 8-pixel boundaries, saving 4 times. Table 1 shows the relationship between the number of boundaries to be filtered and the block size. Therefore, in high-resolution video, when there are many large block selection probabilities, this method can significantly reduce the number of filtering boundaries, thereby reducing the complexity of the codec.

Table 1. The relationship between the number of required filtering borders and the size of the block size

block size Filtering edge count saving 64x64 8 times 32x32 4 times 16x16 2 times 8x8 1 times

Second, this scheme presets filter segmentation blocks (such as 16x16 block size, maximum coding unit block size or image frame, which can be described in the code stream). Before filtering, the image is first divided according to the filter segmentation block, and then scanned according to Sequentially sequentially filter the vertical boundary and then the horizontal boundary of each filter segmentation block, which can facilitate and unify the filter design, improve adaptability and reduce the design complexity of the hardware circuit.

Third, the new filter threshold coefficient table of the present invention is shown in Attached Table 1 and Filter Usage Attached Table 2. At the same time, the filtering direction can be selected according to the mode information and the local image characteristics of the block boundary, instead of being fixed as a vertical block. boundary, which can better improve coding efficiency and subjective quality.

Those skilled in the art can also understand that various illustrative logical blocks (illustrative logical blocks), units, and steps listed in the embodiments of the present invention can be implemented by electronic hardware, computer software, or a combination of both. To clearly demonstrate the interchangeability of hardware and software, the various illustrative components, units and steps above have generically described their functions. Whether such functions are implemented by hardware or software depends on the specific application and overall system design requirements. Those skilled in the art may use various methods to implement the described functions for each specific application, but such implementation should not be understood as exceeding the protection scope of the embodiments of the present invention.

The various illustrative logic blocks or units described in the embodiments of the present invention can be implemented by general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to implement or operate the described functions. The general-purpose processor may be a microprocessor, and optionally, the general-purpose processor may also be any conventional processor, controller, microcontroller or state machine. A processor may also be implemented by a combination of computing devices, such as a digital signal processor and a microprocessor, multiple microprocessors, one or more microprocessors combined with a digital signal processor core, or any other similar configuration to accomplish.

The steps of the method or algorithm described in the embodiments of the present invention may be directly embedded in hardware, a software module executed by a processor, or a combination of both. The software modules may be stored in RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, removable disk, CD-ROM or any other storage medium in the art. Exemplarily, the storage medium can be connected to the processor, so that the processor can read information from the storage medium, and can write information to the storage medium. Optionally, the storage medium can also be integrated into the processor. The processor and the storage medium can be set in the ASIC, and the ASIC can be set in the user terminal. Optionally, the processor and the storage medium may also be set in different components in the user terminal.

In one or more exemplary designs, the above functions described in the embodiments of the present invention may be implemented in hardware, software, firmware or any combination of the three. If implemented in software, the functions can be stored on or transmitted over as one or more instructions or code on a computer-readable medium. Computer-readable media includes computer storage media and communication media that facilitate transfer of a computer program from one place to another. Storage media may be any available media that can be accessed by a general purpose or special computer. For example, such computer-readable media may include, but are not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other device that can be used to carry or store instructions or data structures and Other medium of program code in a form readable by a general-purpose or special-purpose computer, or a general-purpose or special-purpose processor. In addition, any connection is properly defined as a computer-readable medium, for example, if the software is transmitted from a web site, server, or other remote source via a coaxial cable, fiber optic computer, twisted pair, digital subscriber line (DSL) Or transmitted by wireless means such as infrared, wireless and microwave are also included in the definition of computer readable media. Disks and discs include compact discs, laser discs, optical discs, DVDs, floppy discs, and Blu-ray discs. Disks usually reproduce data magnetically, while discs usually reproduce data optically with lasers. Combinations of the above can also be contained on a computer readable medium.

Attached Table 1 The relationship between the block boundary thresholds α and β of the present invention and Index

Attached table 2. Filter parameter design used in the present invention

L3 L2 L1 L0 R0 R1 R2 R3 BS=4 L0 5 4 9 7 4 3 L1 4 3 6 3 L2 1 3 2 1 1 BS=3 L0 1 4 6 4 1 L1 3 8 4 1 BS=2 L0 3 10 3 BS=1 L0 3 1

Attached table 3. AVS original filter parameter design

L3 L2 L1 L0 R0 R1 R2 R3 BS=4 L0 9 9 8 6 L1 6 7 3 L2 3 4 1 BS=3 L0 1 4 6 4 1 L1 3 8 4 1 BS=2 L0 3 10 3 BS=1 L0 3 1

Application of specific embodiments herein has explained the principle of the present invention and implementation mode, and the description of the above embodiments is only used to help understand the method of the present invention and its core idea; Simultaneously, for those of ordinary skill in the art, according to the present invention Thoughts, there will be changes in specific implementation methods and application ranges. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims (7)

1. A method for loop filtering at the video codec end, characterized in that: The loop filtering method comprises the steps of: Step 1) Use the codec end block division information to obtain the image block division structure, and loop filtering is performed on the boundary of the image block. If the coding residual of the coding block is 0, (that is, CBP=0), only the coding block Filter with the boundary of the predicted block; the filtering process is a predefined filter segment block, which divides the image into filter segment unit blocks, and the size of the filter segment block can be predefined or specified in the code stream. In order to unify the filter design, the filtering order is sequential For each filtering segment block, first the vertical boundary and then the horizontal boundary, that is, first filter all the vertical boundaries in the filtering segment block and then filter the horizontal boundary; Step 2) If the boundary includes the coding block, prediction block and transformation block boundary, use the mode, motion information, quantization parameters and residual to make boundary-level filtering decisions, 2.1. If the decision needs to filter the boundary, proceed to the next pixel-level filtering process; 2.2. Otherwise, skip the boundary filtering; make a pixel-level filtering decision at the block boundary that needs to be filtered to determine whether each pixel row needs filtering and the filtering strength (BS); Step 3) Finally, pixel-level filtering is performed. The filtering direction can be perpendicular to the boundary or at other angles according to the local pixel characteristics or block mode and the boundary. The block structure involved in image filtering includes but is not limited to coding blocks, prediction blocks and transformation blocks. ;Filter segmentation blocks are used to facilitate filtering and other operations, and the image is divided in a uniform size. For example, the image can be divided according to 16x16 size filter segmentation blocks. 2 . The method for loop filtering at the video codec end according to claim 1 , wherein the step 1) further comprises: the image block includes a coding block, a prediction block and a transformation block. 3 . 3. A method for loop filtering at the video codec end according to claim 1, characterized in that said step 1) further comprises: the boundary where the loop filtering occurs is at the boundary of the image block, when the coding block When the coding residual is zero (that is, CBP=0), filtering is only performed at the boundary of the coding block and the prediction block, and the minimum size of the filtering (such as 8x8 block size) is defined at the same time; the size of the coding block includes from 64x64 to 8x8, According to the RDO selection, quadtree iterative recursive division is used, which is the basic unit of encoding processing; the prediction block is further divided according to the prediction characteristics on the basis of the coding block; the transformation block is also based on the coding block based on the residual. Further division of distribution characteristics. 4. A method for loop filtering at the video coding and decoding end according to claim 1, characterized in that said step 1) further comprises: dividing the whole frame of image into filter segmentation unit blocks, which can be predefined or in code The stream specifies the size of the filtering block, and filters each filtering block in turn according to the scan, first checks the vertical boundary, then the horizontal boundary, and checks whether the boundary includes the boundary of the coding block, prediction block and transformation block, and if so, filters the boundary boundary, otherwise skip filtering the boundary. 5. A method for loop filtering at the video codec end according to claim 1, characterized in that said step 2) further comprises: according to the mode, motion information, quantization parameters and residuals of the blocks on both sides of the filtering boundary, etc. To determine whether the current block boundary needs to be filtered, if the boundary-level decision needs to filter the boundary, then the next pixel-level filtering process is performed, otherwise the boundary filtering is skipped, if the blocks P and Q on both sides of the block boundary have intra prediction mode, then Filtering needs to be performed; if the P and Q blocks are both inter-frame prediction blocks of the P frame, their residuals are 0, the reference frame is the same and each component of the motion vector is less than one pixel, then it is decided that the block boundary does not need to be filtered, otherwise all need filtering. 6. A method for loop filtering at the video codec end according to claim 1, characterized in that said step 2) further comprises: each line of pixels at the border of the block that needs to be filtered in the filtering decision, using the jump of pixels on both sides of the border The corresponding value of the threshold table corresponding to the index obtained by changing the average QP value of the two blocks is compared to select the filtering strength; the pixel-level filtering strength ranges from 0 to 4, BS=0 means no filtering is required, BS=4 means relatively In flat areas, strong smoothing filtering is required. 7. A method for loop filtering at the video codec end according to claim 1, characterized in that said step 3) further comprises: selecting the corresponding FIR filter for the position of the pixel, and the filtering direction can be perpendicular to the boundary or Other angles can be formed according to local pixel characteristics or block modes and boundaries; FIR filters corresponding to different filter strengths and different pixel positions can be used; at the same time, the texture characteristics of the local image and the mode information of the blocks on both sides of the boundary can be used to determine whether the filtering method is vertical at or at an angle to the boundary.

Images