CN107040778A - Loop filtering method and loop filtering device - Google Patents

Abstract

The invention provides a loop filter processing device or method. And performing sample offset compensation processing on the pixels subjected to the deblocking filtering processing in the current image unit according to one or more sample offset compensation parameters. All or some of the pixels within the boundary of the sample offset compensation parameter of the current image unit share the same sample offset compensation parameter. The vertical sample offset compensation parameter boundary of the current image unit is shifted left by xs lines from the vertical edge of the current image unit, and the horizontal sample offset compensation parameter boundary of the current image unit is shifted up by ys lines from the horizontal edge of the current image unit. To reduce the line buffer requirement, xs is larger than m, m is the number of pixels on each side of the horizontal side line for deblocking filtering correction, and ys is larger than or equal to 0. The loop filtering method and the loop filtering device provided by the invention can improve the hardware efficiency.

Description

Loop filtering method and loop filtering device

technical field

The present invention relates to a video coding system, in particular to video coding combined with a sample offset compensation (Sample Adaptive Offset, abbreviated SAO) and a sample filter compensation (Adaptive Loop Filter, abbreviated ALF) virtual boundary (virtual boundary) system, the present invention relates to reducing line buffers for sample offset offset (SAO) and sample filter offset (ALF).

Background technique

Motion estimation is an effective inter-frame coding technique for utilizing temporal redundancy in video sequences. Inter-frame coding with motion compensation has been widely used in various international video coding standards. Motion estimation applied in various coding standards is usually block-based, and the motion information (such as coding mode and motion vector) used corresponds to each macro block (macroblock) or similar block structure. In addition, Intra-coding is also adaptively applied, where a picture is processed without reference to other pictures. Inter-predicted or Intra-predicted residues are usually further subjected to transformation, quantization, and entropy coding to generate a compressed video bitstream. During encoding (especially in quantization procedures), coding artifacts occur. To reduce coding impairments, newer coding systems require additional processing on the reconstructed video to improve picture quality. The additional processing is often designed as an in-loop operation, so that the encoder and decoder can derive the same reference picture, improving system performance.

FIG. 1A depicts a schematic diagram of an adaptive inter/intra video coding system including in-loop processing. For inter prediction (Inter prediction), the Motion Estimation (ME)/Motion Compensation (MC) 112 is used to provide prediction data based on other video data of a single frame or multiple frames. The switcher 114 selects the intra prediction (Intra Prediction) 110 or the inter prediction data, and the selected prediction data is supplied to the adder 116 to form prediction errors, also known as residues. The prediction error is then processed by a transformer (Transformation, abbreviated T) 118 followed by a quantizer (Quantization, abbreviated Q) 120 . The converted and quantized residual is then encoded by an entropy encoder (Entropy Encoder) 122 to form a video bitstream corresponding to the compressed video data. The bitstream associated with the transform coefficients is then filled with side information (eg, motion, mode, or other information related to the image region). Side information can also be used for entropy coding to reduce bandwidth requirements. Therefore, side information related data may be supplied to the entropy encoder 122 as shown in FIG. 1A . When using the inter prediction mode, a single reference picture or multiple reference pictures must also be reconstructed at the encoder. Therefore, the converted and quantized residual is processed by an inverse quantization (Inverse Quantization, IQ for short) 124 and an inverse transformation (Inverse Transformation, IT for short) 126 to recover the residual. The restored residual can be added back to the prediction data 136 at a reconstruction (REC) 128 to reconstruct video data. The reconstructed video data can be stored in a reference picture buffer (Reference Picture Biffer) 134 and used to predict other frames.

As shown in FIG. 1A, the received video data undergoes a series of processing in the encoding system. The reconstructed video data from the reconstructor 128 may have various impairments due to a series of processes. Therefore, before the reconstructed video data is stored in the reference picture buffer 134, various in-loop processes are performed to improve the video quality. In the development of the High Efficiency Video Coding (abbreviated HEVC) standard, a deblocking filter (Deblocking Filter, abbreviated DF) 130, a sample offset compensation (Sample Adaptive Offset, abbreviated SAO) 131, and a sample filter have been developed. Compensation (Adaptive Loop Filter, ALF for short) 132 to improve image quality. In-loop filter information may need to be incorporated into the bitstream so that the decoder can properly recover the required information. Therefore, the in-loop filter information from sample offset offset (SAO) and sample filter offset (ALF) will be supplied to entropy encoder 122 for inclusion in the bitstream. In Fig. 1A, the deblocking filter 130 is firstly applied to the reconstructed video, the sample offset compensation (SAO) 131 is then applied to the deblocking filter (DF) processed video, and the sample filter compensation (ALF) 132 is applied again Applied to video after Sample Offset Offset (SAO) processing. However, the order of deblocking filter (DF), sample offset compensation (SAO), and sample filter compensation (ALF) is adjustable. The system shown in FIG. 1A can correspond to the High Efficiency Video Coding (HEVC) system (except for Sample Filter Compensation (ALF)), or the video coding standard AVS2 (this is a video and audio coding standard developed by a Chinese team). Sample filter compensation (ALF) has been evaluated in the development of high-efficiency video coding (HEVC), but sample filter compensation (ALF) has not been adopted in current high-efficiency video coding (HEVC).

FIG. 1B is a system block diagram corresponding to a video decoder including a deblocking filter (DF), sample offset compensation (SAO) and sample filter compensation (ALF). Since the encoder may include a local decoder for video data reconstruction, some decoder components (except entropy decoder 142) are already used in the encoder. In addition, motion compensation 144 is still required at the decoder side. A switch 146 selects either inter prediction or intra prediction, and the selected prediction data is used in a reconstructor (REC) 128 for combination with the restored residual. In addition to performing entropy decoding on the compressed video data, the entropy decoder 142 is also responsible for entropy decoding of side information to provide side information for corresponding blocks. For example, the intra mode information is supplied to the intra prediction buffer 111, the inter mode information is supplied to the motion compensation 144, the adaptive offset information is supplied to the sample offset compensation (SAO) 131, the sample filter compensation The information is supplied to sample filter compensation (ALF) 132 and the residual is supplied to inverse quantizer (IQ) 124 . The residual is processed by inverse quantization (IQ) 124, inverse transformation (IT) 126 and subsequent reconstruction procedures to reconstruct the video data. Once again, the reconstructed video data provided by the reconstructor (REC) 128 is after a series of processing including inverse quantization (IQ) 124 and inverse transformation (IT) 126 as shown in FIG. 1B , there is an intensity shift (intensity shift) . The reconstructed video data is further processed by deblocking (DF) 130 , sample offset compensation (SAO) 131 , and sample filter compensation (ALF) 132 .

The encoding procedure of High Efficiency Video Coding (HEVC) is implemented according to a largest coding unit (Logic Coding Unit, LCU for short, also known as a coding tree unit (CTU for short)). The largest coding unit is adaptively partitioned into a plurality of coding units using a quadtree. In High Efficiency Video Coding (HEVC), Deblocking Filter (DF) works on 8x8 blocks. For each 8x8 block, horizontal filtering across vertical block boundaries is performed first, followed by vertical filtering across horizontal block boundaries. 2A illustrates deblocking filtering (DF) processing of a high-efficiency video coding (HEVC) luma (luma) component. Four boundary pixels need to be considered on both sides of the block boundary 210 . A boundary may correspond to a vertical boundary or a horizontal boundary. The boundary pixels are labeled q0, q1, q2 and q3 and p0, p1, p2 and p3. Two pixels, q0 and p0, are adjacent to the boundary. During processing of luma block boundaries, four pixels on each side are used for filter parameter derivation, and up to three pixels on each side (ie, p0, p1, p2 or q0, q1, q2) can be filtered. Regarding horizontal filtering across vertical block boundaries, unfiltered reconstructed pixels are used for filter parameter derivation and also serve as raw pixels for filtering. Regarding vertical filtering across horizontal block boundaries, deblocking filtered (DF) processed intermediate pixels (ie, horizontally filtered pixels) are used for filter parameter derivation and are also used as raw pixels for filtering. In the deblocking filtering (DF) process of the high-efficiency video coding (HEVC) chroma (chroma) component, two boundary pixels are used on each side of the block boundary, and only one pixel on each side (i.e., p0 or q0) is modified .

FIG. 2B depicts deblocking filtering (DF) processing of the video coding standard AVS2 luma component, involving three boundary pixels on each side of a block boundary 220 . The border pixels are labeled q0, q1, q2 and p0, p1 and p2, and two pixels q0 and p0 are next to the border. As for deblocking filtering (DF) processing of chroma block boundaries, two pixels on each side are used for filter parameter derivation. Corresponding to the video coding standard AVS2, the deblocking filtering (DF) process can correct all the boundary pixels involved. In other words, three luma pixels and two chroma pixels on each side of the block boundary can be adjusted.

The types of sample offset compensation (SAO) based on High Efficiency Video Coding (HEVC) and the video coding standard AVS2 are shown in Figure 3. There are four types of sample offset compensation (SAO) corresponding to 0 degrees, 90 degrees, and 135 degrees. And four directions of 45 degrees. Sample Offset Offset (SAO) performs per-pixel in-loop filtering on each pixel. The sample offset offset (SAO) parameter is updated corresponding to each largest coding unit (LCU) or coding tree unit (CTU). Corresponding to the direction type of sample offset compensation (SAO), first implement pixel classification, such as classifying pixels into multiple groups (groups, also called categories or classes) according to the classification status in Table 1 . After classification, each reconstructed and deblock filtered (DF) pixel is compensated by an offset value based on the selected orientation type and classification result.

Table 1

type situation 1 C< two adjacent pixels 2 C<one adjacent pixel &&C==another adjacent pixel 3 C> one adjacent pixel && C==another adjacent pixel 4 C> two adjacent pixels 0 None of the above

As shown in Table 1, the implementation of the sample offset offset (SAO) classification situation can compare the center pixel (C) separately from the two adjacent pixels. Classification checks whether the central pixel is greater than, less than, or equal to the corresponding neighboring pixel. The three kinds of comparison results can be represented by two bits (2-bit) data.

Sample offset compensation (SAO) parameters (eg, pixel offset value (pixel offset) and sample offset compensation type (SAO type)) can be adaptively determined corresponding to each coding tree unit (CTU). For High Efficiency Video Coding (HEVC), the Sample Offset Offset (SAO) parameter boundary is the same as the Coding Tree Unit (CTU). The sample offset offset (SAO) processing of all pixels within the parameter boundary shares the same sample offset offset (SAO) type and offset value. Because sample offset compensation (SAO) is applied to pixels after deblocking filtering (DF), the sample offset compensation process (SAO process) of the current coding tree unit (CTU) must wait until the deblocking of the current coding tree unit (CTU) It can only be performed after block filtering (DF) processing is completed. However, pixels around a CTU boundary cannot be deblocking filtered (DF) until the reconstructed video data is available next to the CTU boundary on the other side of the CTU boundary . Based on such data dependencies, the video coding standard AVS2 uses offset Sample Offset Offset (SAO) parameter boundaries. Fig. 4 depicts an example of shifting of parameter boundaries of Sample Offset Offset (SAO) according to the video coding standard AVS2. For High Efficiency Video Coding (HEVC), embodiments 410 of corresponding Sample Offset Offset (SAO) parameter boundaries correspond to Coding Tree Unit (CTU) boundaries. For the video coding standard AVS2, the corresponding sample offset offset (SAO) parameter boundary 420 is shifted left by xS and upward by yS corresponding to the coding tree unit (CTU) boundary. Furthermore, in the video coding standard AVS2, xS=4 and yS=4.

Sample Filter Compensation (ALF) 132 is a video coding tool used to improve picture quality. Sample Filter Compensation (ALF) has been evaluated during the development phase of High Efficiency Video Coding (HEVC). However, the sample filter compensation (ALF) is not used in the current high-efficiency video coding (HEVC) standard, but has been incorporated into the video coding standard AVS2. In particular, the 17-tap symmetrical sample filter compensation (ALF) used in the video coding standard AVS2 as shown in FIG. 5 . The 17-tap symmetrical sample filter compensation (ALF) means that the filtering operation of the current pixel may require the data of the lower three lines (lines). If the rows are from another CTU (specifically, the CTU located in a subsequent CTU row), the Sample Filter Compensation (ALF) process must be delayed until the subsequent related data. The above characteristics mean that a line buffer (line buffer) needs to be configured to instantaneously store the relevant data of the current coding tree unit (CTU) for subsequent processing. In order to deal with the problem of data dependence, the video coding standard AVS2 adopts an ALF virtual boundary to restrict ALF processing from crossing the virtual boundary. 6 depicts an example of the sample filter compensation (ALF) virtual boundary of the luma component according to the video coding standard AVS2, which depicts the sample filter compensation (ALF) for selected pixels (such as pixels a, b, c and d). )deal with. Line 610 is a coding tree unit (CTU) boundary between coding tree unit (CTU) X and coding tree unit (CTU) Y. Row 620 is the sample filter compensation (ALF) virtual boundary (i.e. yC-4) of luminance, which is located four lines above the coding tree unit (CTU) boundary (i.e. yC) according to the specification of the video coding standard AVS2 (i.e. at yC-4). 4) place. Regarding the chroma component, according to the video coding standard AVS2 (see: Information Technology–Advanced Media Coding Part2: Video Final Committee Draft, Audio and Video Coding Standard Workgroup of China, Feb.7, 2015, Document: N2120.D3), sample filter compensation The (ALF) virtual boundary is located at a distance of 3 lines (ie, at yC-3) above the coding tree unit (CTU) boundary. The sample filter compensation (ALF) processing of pixels a, b and c is performed in the coding tree unit (CTU) X processing stage. Furthermore, the sample filter compensation (ALF) process for pixels a, b and c only uses information above the virtual boundary. As for the pixel d below the virtual boundary, its sample filter compensation (ALF) processing is implemented in the coding tree unit (CTU) Y processing stage, and only the information below the virtual boundary is used. The use of virtual boundaries will suppress data dependencies and reduce the capacity requirements of the row buffer.

As mentioned above, the deblocking filter (DF), sample offset offset (SAO) and sample filter offset (ALF) procedures involve adjacent data. In High Efficiency Video Coding (HEVC) and Video Coding Standard (AVS2), Coding Tree Unit (CTU) has been used as the unit of encoding process. When deblocking filter (DF), sample offset offset (SAO), and sample filter offset (ALF) processes are performed across coding tree unit (CTU) boundaries, data dependencies must be carefully managed to reduce line buffer requirements . Since the deblocking filter (DF), sample offset offset (SAO) and sample filter offset (ALF) processes are sequentially implemented in each coding tree unit (CTU), the corresponding hardware implementation must be designed as a pipeline. Figure 7 depicts the data dependencies of the Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) processes for the video coding standard AVS2 decoder. A processing sequence 700 using CTUs is shown in FIG. 7 , and the CTU boundary between CTU X and CTU Y is labeled 705 . As shown in FIG. 7 , the reconstructed video from reconstruction block 710 is processed by deblocking filter (DF) 720 , sample offset compensation (SAO) 730 and sample filter compensation (ALF) 740 . The output of sample filter compensation (ALF) 740 is stored in the decoded frame buffer.

The corresponding processing states of the deblocking filter (DF) 720 , sample offset offset (SAO) 730 , and sample filter offset (ALF) 740 processes are numbered 725 , 735 , and 745 , respectively. Graph 725 shows the deblocking filtering (DF) processing status at the end of the deblocking filtering (DF) processing stage of coding tree unit (CTU) X. Luma pixels above row 722 and chroma pixels above row 724 have deblocking filtering (DF) done. Because the pixels on the other side of the block boundary (i.e., below the coding tree unit (CTU) boundary 705) are not yet available, the luma pixels below row 722 and the chrominance pixels below row 724 cannot be located in the coding tree unit (CTU) X. are processed in the Block Filtering (DF) processing stage. Graph 735 shows the state of the sample offset compensation (SAO) processing at the end of the sample offset compensation (SAO) processing phase of coding tree unit (CTU) X. The luma pixels above row 732 and the chroma pixels above row 734 have sample offset compensation (SAO) done, where row 732 is aligned with row 734 . Graph 745 shows the sample filter compensation (ALF) processing status at the end of the sample filter compensation (ALF) processing stage of coding tree unit (CTU) X. Similarly, the luma pixels below row 732 and the row Chroma pixels below 734 cannot be used for sample offset compensation (SAO) of coding tree unit (CTU) X. Luma pixels above row 742 (Ample Filtered (ALF) Virtual Boundary for Luma) are sample filtered (ALF) processed based on the video coding standard AVS2 draft. Chroma pixels above row 744 (the ALF virtual boundary for chroma) may be ALF processed. However, sample filter compensation (ALF) of chroma components cannot be performed on chroma lines A to D in the coding tree unit (CTU) X processing stage. For example, an sample filter compensation (ALF) process for pixel 746 would use pixel 748 . Because the chroma pixel 748 is located below the SAO boundary 734 , the chroma pixel 748 has not been SAO processed in the coding tree unit (CTU) X processing stage. Therefore, although located above the ALF virtual boundary of chroma, the chroma pixel 746 is not ALF-processed. Therefore, the six lines above pixel 748 (i.e., above line D) that have undergone sample offset compensation (SAO) processing must be stored in a buffer for later processing of lines A to D in the coding tree unit (CTU) Y processing stage. Used during sample filter compensation (ALF) processing. Among them, the three lines above line A have been subjected to sample filter compensation (ALF) at the coding tree unit (CTU) X processing stage, but the sample filter compensation (ALF) processing on line A is also required.

In terms of hardware implementation, chrominance samples with a picture width of six lines must be stored in a line buffer, which is generally implemented using an embedded memory, and such an application may require high chip costs. Therefore, it is desired to develop a method and an apparatus for reducing loop filter processing (such as deblocking filter (DF), sample value offset compensation (SAO), sample filter compensation (ALF) processing, other loop filter processing or combination) the number of line buffers required. In addition, for different sample value offset compensation (SAO) parameter boundaries, the system will switch between different sample value offset compensation (SAO) parameters. This will increase system complexity and power consumption. Therefore, it is desirable to develop loop filtering processes (such as deblocking filter (DF), sample offset compensation (SAO), sample filter compensation (ALF) processing other loop filtering procedures or combinations thereof) with appropriate system parameter design , to reduce line buffer requirements, system complexity, and system power consumption, or to obtain a combination of any of the above improvements. In another point of view, it is desirable in the art to develop a method and apparatus for performing deblocking filter (DF), sample offset compensation (SAO), sample filter compensation (ALF), Loop filter processing of other loop filter procedures or combinations thereof for use in video coding systems incorporating such loop filter processing.

Contents of the invention

The invention discloses a method and a device for loop filtering processing of reconstructed video data. In order to reduce the computational complexity of sample offset compensation (SAO) parameter switching and reduce the demand for line buffers, the embodiments of the present invention make horizontal and vertical displacements for the sample value offset compensation (SAO) parameter boundary according to the corresponding target . According to an embodiment of the present invention, deblocking filtering (DF) processing is first performed on the reconstructed pixels. The deblocking filtering (DF) process modifies up to m pixels on each side of a horizontal edge corresponding to an image unit boundary between two image units. According to one or more sample value offset compensation (SAO) parameters, sample value offset compensation (SAO) processing is performed on the pixels processed by deblocking filter (DF) in the current image unit. All or part of the pixels within the boundary of the SAO parameters of the current image unit share the same one or more SAO parameters. The vertical sample offset compensation parameter (SAO) boundary of the current image unit is shifted to the left by xs lines from the vertical edge of the current image unit, and the horizontal sample offset compensation (SAO) parameter boundary of the current image unit is changed from the horizontal edge of the current image unit Edge shifted up by ys rows. According to one or more spatial loop filter parameters, spatial loop filtering is performed on the pixels above the narrowing boundary of the spatial loop filter in the current image unit that have been processed by Sample Offset Compensation (SAO). The spatial loop filter confinement boundary of the current image unit is shifted upwards by yv lines from the bottom edge of the current image unit. In order to reduce the line buffer requirements, m, xs, ys and yv are positive integers, xs is greater than m, ys is greater than or equal to 0, and ys is less than yv, and yv is set according to m.

Each image unit may correspond to a coding tree unit (CTU). Spatial loop filter processing may correspond to sample filter compensation (ALF) processing.

If the above-mentioned reconstructed video data includes a luma component and a chroma component, the above-mentioned deblocking filter processing, sample value offset compensation processing, and spatial loop filter processing are performed in a luma component and a chroma component, and the above m The values are labeled M and N, the above-mentioned xs used respectively are labeled xS and xSC, and the above-mentioned ys used respectively are labeled yS and ySC. Also, the above-mentioned yv used respectively are labeled as yV and yVC. In one embodiment, yS and ySC are equal to zero. yV may be greater than M and yVC may be greater than N. For example, yV is equal to (M+1) and yVC is equal to (N+1). In one embodiment, M is equal to 3 and N is equal to 2.

In another embodiment, yS is equal to ySC, yV is equal to yVC, and yVC is greater than MAX(M,N). For example, yV and yVC are equal to MAX(M, N)+1 and yS and ySC can be integers ranging from 0 to MAX(M, N). In one embodiment, M is equal to 3 and N is equal to 2.

In another embodiment, the sign data obtained by comparing the current pixel of the current row processed in the current processing stage of the current image unit with the adjacent pixels of the adjacent row processed in the subsequent processing stage are stored, and the above sign data Corresponds to "greater than", "less than", or "equal to". The above symbol data can be stored using 2 bits.

The loop filtering method and the loop filtering device proposed by the present invention can improve hardware efficiency.

Description of drawings

Figure 1A depicts an adaptive inter/intra video coding system in which a loop filter including a deblocking filter (DF), a sample offset offset (SAO) and a sample filter offset ( ALF) is used to process the reconstructed video data;

FIG. 1B depicts a system block diagram corresponding to a video decoder, which includes a deblocking filter (DF), sample offset compensation (SAO) and sample filter compensation (ALF);

2A depicts pixels on both sides of a block boundary involved in deblocking filtering (DF) processing of luma components according to High Efficiency Video Coding (HEVC);

FIG. 2B depicts pixels on both sides of the block boundary involved in deblocking filtering (DF) processing of the luma component according to the video coding standard AVS2;

Figure 3 describes the pixel classification based on the 3x3 window, there are four pointing forms, corresponding to 0 degrees, 90 degrees, 135 degrees and 45 degrees;

Fig. 4 has described the sample offset compensation (SAO) boundary displacement according to video coding standard AVS2;

Fig. 5 has described the 17-tap (17-tap) symmetrical sample filter compensation (ALF) that is used for video coding standard AVS2;

Figure 6 depicts the sample filter compensation (ALF) virtual boundary of the luma component according to the video coding standard AVS2;

Figure 7 depicts the data dependencies associated with the Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) processes for a decoder of the video coding standard AVS2;

Fig. 8 describes the processing state of the deblocking filter (DF), sample value offset compensation (SAO) and sample filter compensation (ALF) processing, wherein the boundary parameters related to the loop filter are indicated;

FIG. 9 depicts sample offset compensation (SAO) processing with different horizontal sample offset offset (SAO) parameter boundaries and horizontal sample offset compensation (SAO) processing boundaries according to an embodiment of the present invention;

Fig. 10 depicts the processing state of the Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) processes at the end of the current Coding Tree Unit (CTU) processing stage according to the first embodiment of the present invention ;

Fig. 11 depicts the processing state of the Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) processes at the end of the current Coding Tree Unit (CTU) processing stage according to the second embodiment of the present invention ;

Figure 12 depicts a flow diagram of an encoding system incorporating an embodiment of the present invention that aligns loop filter correlation boundaries to reduce line buffer requirements.

detailed description

The enumerated preferred embodiments of the present invention are described below. The purpose of the following description is to introduce the basic concept of the present invention, and is not intended to limit the content of the present invention. The scope of protection of the present invention should be based on the scope determined by the claims.

In order to facilitate the discussion of data dependencies between different loop processing stages, this application introduces a loop filter with boundary parameter correlation. Figure 8 reproduces the relevant processing states of the deblocking filter (Deblocking Filter, DF), sample offset compensation (Sample Adaptive Offset, SAO) and sample filter compensation (Adaptive Loop Filter, ALF) processing in Figure 7, and the loop Filter-related boundary parameters are marked therein. The processing states for the Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) are numbered 825, 835 and 845, respectively. Graph 825 shows the DF processing status of the coding tree unit (CTU) X at the end of the DF processing stage. Luma pixels above row 822 (ie, luma DF boundary) and chroma pixels above row 824 (ie, chroma DF boundary) are deblock filtered (DF). The maximum number of luma pixels is M, and the maximum number of chroma pixels is N, on each side of the block boundary updated in the deblocking filtering (DF) process. In FIG. 8, the coding tree unit (CTU) boundary 805 is also a block boundary.

Graph 835 shows the SAO processing status of CTU X at the end of the SAO processing stage. Luma pixels above row 832 (i.e., the sample offset offset (SAO) boundary for luma) and chrominance pixels above row 834 (i.e., the sample offset offset (SAO) boundary for chroma) are sample offset compensated ( SAO) processing. Line 832 is aligned with line 834 based on the sample offset compensation (SAO) parameter boundary displacement technique proposed by the video coding standard AVS2. In order to avoid sample offset compensation (SAO) parameter switching in each coding tree unit (CTU) processing stage, the sample value offset compensation (SAO) parameter boundary is at the luma component displacement (xS, yS), and at the color Degree component displacement (xSC, ySC). In other words, corresponding to the coding tree unit (CTU) whose upper left point is (xC, yC), the upper boundary of the sample value offset compensation (SAO) parameter corresponding to the luma component is shifted to (yC-yS), and corresponding to the color The upper boundary of the sample offset compensation (SAO) parameter of the degree component is shifted to (yC-ySC) (as shown in FIG. 8 ). Similarly, the Sample Offset Offset (SAO) parameter boundary displacement is also implemented in the x direction. Initially, under the High Efficiency Video Coding (HEVC) specification, the Sample Offset Offset (SAO) parameter is determined according to the Coding Tree Unit (CTU). In the prior art, in order to reduce the computational complexity of sample offset compensation (SAO) parameter switching, the sample offset compensation (SAO) parameters are based on deblocking filter (DF) processed and ready for sample offset compensation (SAO) processing of these pixel decisions. That is, prior art SAO parameter boundaries are equivalent to deblocking filtered (DF) processed pixel boundaries ready for SAO processing. A deblocking filter (DF) processed pixel boundary ready for SAO processing is also called a SAO processing boundary.

Graph 845 shows the ALF processing status of the coding tree unit (CTU) X at the end of the ALF processing stage. Luma pixels above row 842 (ie, the ALF virtual boundary for luma) are ALF processed. Chroma pixels above row 844 (ie, the ALF virtual boundary for chroma) may be ALF processed. However, chroma row D cannot perform sample filter compensation (ALF) processing of chroma components at the coding tree unit (CTU) X processing stage. The sample filter compensation (ALF) virtual boundary corresponding to the luma component is (yC-yV), and the sample filter compensation (ALF) virtual boundary corresponding to the chroma component is (yC-yVC). Among them, yV and yVC respectively correspond to the boundary vertical displacement of luma and chrominance components. For the video coding standard AVS2 draft, the number of boundary pixels (ie, M and N) of the luma and chroma components to be updated are 3 and 2, respectively. The Sample Offset Offset (SAO) parameter Boundary Vertical Offset is set to 4 for both luma and chroma components. On the other hand, the vertical shifts (ie, yV and yVC) of the ALF virtual boundaries of the luma and chrominance components are set to 4 and 3, respectively.

In order to simultaneously reduce the line buffer size requirements and reduce the computational complexity of sample offset compensation (SAO) parameter switching in the coding tree unit (CTU) processing stage, the present application discloses a sample offset compensation (SAO) parameter boundary Setting skills, according to each target, the horizontal and vertical directions of displacement of the sample offset compensation (SAO) parameter boundary, and there are different sample value offset compensation (SAO) parameter boundaries and sample value offset compensation (SAO) processing boundaries . As mentioned above, in the prior art, the sample offset compensation (SAO) parameter boundary and the sample value offset offset (SAO) processing boundary are the same. According to the present invention, the vertical sample offset offset (SAO) parameter boundary remains equal to the sample offset offset (SAO) processing boundary, but the horizontal sample offset offset (SAO) parameter boundary can be different from the sample offset offset (SAO ) handles boundaries. In particular, the Sample Offset Offset (SAO) processing boundary is determined according to the location of the DF-processed pixel data. Figure 9 depicts unequal horizontal sample offset offset (SAO) parameters and processing boundaries according to an embodiment of the present invention. In one embodiment, the horizontal sample offset offset (SAO) parameter boundary 912 is equivalent to the coding tree unit (CTU) horizontal boundary 910 . On the other hand, a horizontal sample offset offset (SAO) processing boundary 920 is located one row above the deblocking boundary 930 . As for the pixel 940 above the horizontal sample offset offset (SAO) processing boundary 920, its sample offset offset (SAO) processing is based on the sample offset offset (SAO) parameter boundary of the coding tree unit (CTU) X ( That is, the Sample Offset Offset (SAO) parameter above it). However, according to the content of the present application, the pixel 950 on row D is also subjected to sample offset compensation (SAO) processing based on the sample value offset compensation (SAO) parameter of the coding tree unit (CTU) X, because the pixel 950 is located in the horizontal sample Value Offset Compensation (SAO) parameter boundary 912 (ie, above). In traditional sample offset compensation (SAO) processing, the pixel 950 on row D is subjected to sample offset compensation (SAO) according to the sample value offset compensation (SAO) parameter of coding tree unit (CTU) Y.

In the above discussion, an image is divided into multiple coding tree units, and each coding tree unit is divided into one or more coding units (CUs). Deblocking Filter (DF), Sample Offset Offset (SAO) and Sample Filter Offset (ALF) processes are performed within block boundaries to reduce artifacts on or near block boundaries. For coding systems in which coding tree units (CTUs) are processed in horizontal raster scan order, deblocking filters (DF) bounded by coding tree units (CTUs, also known as block extents), sample offset compensation ( SAO) and sample filter compensation (ALF) processing will require row buffers to store information across coding tree unit row (CTU row) boundaries. However, the image can also be divided into other image units (such as macroblocks or tiles) for encoding processing. Image unit boundaries face the same line buffer issues as coding tree unit (CTU) boundaries face.

The ALF described above is just an example, and the present invention can be applied to any spatial loop filter. For example, a two-dimensional finite impulse response (FIR) filter with a set of spatial loop filter parameters may be used instead of sample filter compensation (ALF). In order to reduce the line buffer requirements of the spatial loop filter processing, the restricted spatial loop filter boundary can be used to restrict the spatial loop filter processing to only use the sample offset offset (SAO) processed data. For example, the narrowed spatial loop filter boundary may be located a distance y rows above the coding tree unit (CTU) boundary. The spatial loop filter is implemented on the sample offset offset (SAO) processed pixels above the restricted spatial loop filter boundary, and will only use the sample offset offset above the restricted spatial loop filter boundary. The pixels processed by motion compensation (SAO) are used as the input of the spatial loop filter.

FIG. 9 depicts an embodiment of the present invention in which the horizontal sample offset offset (SAO) parameter boundary 912 need not be the same as the coding tree unit (CTU) horizontal boundary 910 . According to the embodiment of the present invention, the horizontal sample offset compensation (SAO) parameter boundary 912 can be designed in any row from the horizontal boundary 910 of the coding tree unit (CTU) to the next row of the horizontal sample offset compensation (SAO) processing boundary 920 . For the pixels below the horizontal sample offset compensation (SAO) processing boundary 920 in the coding tree unit (CTU) X, because the sample value offset compensation (SAO) operation of the pixels in this area is based on the Sample value offset compensation (SAO) parameter information, so the sample value offset compensation (SAO) parameter information required for the sample value offset compensation (SAO) operation needs to be buffered and stored. However, in High Efficiency Video Coding (HEVC) and video coding standard AVS2, Sample Offset Offset (SAO) parameters can be coded as merge-up or merge-left syntax, and The sample offset offset (SAO) parameters of the entire coding tree unit row (CTU row) need to be buffered and stored, so the stored sample offset offset (SAO) parameters can be shared.

If the video data being processed corresponds to color video data, the technology in this case can be applied to the luma component and the chrominance component. In the first embodiment, the vertical displacement yS and ySC of the horizontal sample offset compensation (SAO) parameter boundary and the vertical displacement yV and yVC of the sample filter compensation (ALF) virtual boundary are determined by formulas (1) and (2) :

0≤yS<yV=M+1, and (1)

0≤ySC<yVC＝N+1 (2)

The main role of the row buffer requirement is to fulfill the storage requirement for the boundary loop filtering process from one CTU row to the next CTU row. Because an image may be very wide, the corresponding line buffer size may be large. Therefore, it is an object of the present invention to reduce the row buffer requirements of the loop filtering process across a coding tree unit (CTU) boundary between two coding tree unit rows (CTU rows). If the boundary displacement corresponding to the vertical boundary has an effect on the row buffer requirements, it is also very small. The SAO parameter boundary horizontal shifts xS and xSC of the luma and chrominance components remain the same as conventional (ie, xS=M+1 and xSC=N+1). When the system processes pictures in vertical scanning order, the CTU column is processed in the same manner as the CTU row.

FIG. 10 depicts an embodiment according to the invention, where M=3, N=2, yS=3, ySC=2, yV=4 and yVC=3. The processing states of the deblocking filter (DF), sample offset offset (SAO) and sample filter offset (ALF) processes are referenced 1025, 1035 and 1045 in FIG. 10, respectively. The loop processing boundary is partly the same as that shown in Figure 8. In FIG. 10 , the reference numerals in FIG. 8 represent the same loop processing boundaries. Figure 10 indicates the sample offset offset (SAO) processing boundary for the luma and chroma components, the vertical shift (yS') of the sample offset offset (SAO) process for the luma component is 4, and the chroma component The vertical displacement (ySC') of the Sample Offset Offset (SAO) process is 3. As shown in FIG. 10 , SAO parameter boundaries 1032 and 1034 for luma and chroma components are one row below corresponding SAO processing boundaries 1036 and 1038 .

The displacement in the horizontal direction of the sample value offset compensation (SAO) parameter boundary is equal to the sample value offset compensation (SAO) processing boundary. For example, the horizontal displacement xS and xSC of the sample value offset compensation (SAO) parameter boundary of luma and chrominance components is the same as the horizontal displacement of the sample value offset compensation (SAO) processing boundary.

In the second embodiment, the sample value offset compensation (SAO) parameter boundary vertical displacement yS and ySC and the sample filter compensation (ALF) virtual boundary vertical displacement yV and yVC are determined according to formula (3):

0≤yS=ySC<yV=yVC=MAX(M,N)+1 (3)

In other words, the sample offset offset (SAO) parameter boundaries for luma and chroma components are the same, making memory access behavior consistent. The sample filter compensation (ALF) virtual boundaries for luma and chroma components are also the same. Additionally, the Sample Filter Offset (ALF) virtual boundary is located at least one row above the Sample Offset Offset (SAO) parameter boundary.

FIG. 11 depicts an embodiment according to the present invention, where M=3, N=2, yS=ySC=2, yV=VC=4. The processing states of the deblocking filter (DF), sample offset offset (SAO) and sample filter offset (ALF) processes are referenced 1125, 1135 and 1145 in FIG. 11, respectively. The loop processing boundary is partly the same as that shown in Figure 8. In FIG. 11, the same reference numerals as in FIG. 8 denote the same loop processing boundaries. Figure 11 also depicts the sample value offset compensation (SAO) processing boundary for the luma and chroma components, the vertical displacement of the sample value offset compensation (SAO) processing boundary of the luma component and the chrominance component (that is, labeled yS 'with ySC') is 4. As shown in FIG. 11 , SAO parameter boundaries 1132 and 1134 for the luma and chroma components are located two rows below corresponding SAO processing boundaries 1136 and 1138 . The figure further indicates sample filter compensation (ALF) virtual boundaries 1142 and 1044 for luma and chrominance components.

The sample offset offset (SAO) parameter boundary in the horizontal direction is equal to the sample offset offset (SAO) processing boundary in the horizontal direction. For example, the sample value offset compensation (SAO) parameters boundary horizontal displacement xS and xSC of luma and chrominance components can be set as xS=xSC=MAX(M,N)+1.

In sample offset compensation (SAO) processing, pixel data in the current coding tree unit (CTU) processing boundary can be used for later sample offset compensation (SAO) processing. For example, row D of FIG. 9 is processed at the pipeline stage of coding tree unit (CTU) Y. The sample offset offset (SAO) processing of row D will require the data of row C; the row C data is processed in the sample offset offset (SAO) processing stage of coding tree unit (CTU) X. As previously discussed, the sample offset offset (SAO) classification process compares the central pixel data with two individual neighboring pixels. The comparison result determines whether the central pixel is greater than, less than or equal to the selected adjacent pixels. Accordingly, the sample offset compensation (SAO) operation module can preprocess the comparison between row C and row D and store the result (ie, ">", "<" or "="). Because the unaligned sample offset compensation (SAO) parameter boundaries and processing boundaries will cause row D to use the same sample value offset compensation (SAO) parameters for row C to perform sample offset compensation (SAO) processing, the comparison results are in The pipeline-level processing of coding tree unit (CTU) X is obtained when row C is sampled offset offset (SAO) processed.

The comparison result between the pixel row C and the adjacent pixel row D can be represented by two-bit data for each pixel to indicate one of three comparison results. The two-bit symbol data is relatively low in data size compared to full pixel storage, which is typically 8 or more bits. Accordingly, the cost of the line buffer can be greatly reduced.

The loop filter processing boundary design disclosed above can be used to solve the data dependence of the video system (such as the video coding standard AVS2) including deblocking filter (DF), sample value offset compensation (SAO) and sample filter compensation (ALF) processing. The condition causes a large row buffer requirement. The present invention can also be implemented in any high-order video coding system including deblocking filter (DF), sample offset compensation (SAO) and sample filter compensation (ALF) processing.

Table 2 compares the line buffer requirements of the video coding standard AVS2 and the embodiment of the present invention. As stated above, all of the above embodiments require three rows of stored data for each of the luma and chrominance components for deblocking filtering (DF). For sample offset compensation (SAO) processing, all systems need to store line C and line D data for the luma and chrominance components. However, instead of storing row C pixel data, storage requirements can be reduced by storing the comparison result between row C and row D. As mentioned above, each comparison result only needs to be stored with two bits. According to the conventional video coding standard AVS2, the sample offset offset (SAO) results of the six lines will be stored as the sample filter offset (ALF) process of the chroma component. A system incorporating any of the embodiments of the present invention can address the need to store chroma component sample filter compensation (ALF) processing in a six-line buffer. Corresponding to the traditional video coding standard AVS2, the first embodiment and the second embodiment, the total number of line buffers required by the deblocking filter (DF), sample value offset compensation (SAO) and sample filter compensation (ALF) is 16 respectively , 7.5 and 8.5. Both embodiments achieve additional memory compaction by storing the sign of the comparison result involved in the sample offset offset (SAO) process. In other words, the first and second embodiments can reduce 8.5 and 7.5 line buffers, respectively. In the example of Table 2, the sample filter compensation (ALF) chroma virtual boundary of the traditional video coding standard AVS2 is yC-3, and the sample offset compensation (SAO) displacement pixel amount (expressed by the parameter SAO_SHIFT_PIX_NUM) is 4; the first implementation The example corresponds to the content in FIG. 9 ; and the second embodiment corresponds to the content in FIG. 10 .

Form 2

Figure 12 depicts a flow diagram of a video encoding system incorporating an embodiment of the present invention that aligns loop filter correlation boundaries to reduce line buffer requirements. Step 1210, the video system receives reconstructed video data corresponding to one image unit. The reconstructed video data can be taken from memory (e.g., computer memory, a buffer (random access memory (RAM) or dynamic random access memory (DRAM) or other media), or from a processor. Deblocking filtering (DF) Processing is then performed on the reconstructed pixels at step 1220. The deblocking filtering (DF) process corrects up to m pixels on each side of the horizontal boundary corresponding to the image unit boundary between the two image units. Step 1230, based on one or more samples Offset compensation (SAO) parameters, sample value offset compensation (SAO) processes pixels processed by deblocking filter (DF) in the current image unit. Within the boundary of the sample value offset compensation (SAO) parameter of the current image unit All or part of the pixels share the same sample value offset compensation (SAO) parameter. The vertical sample value offset compensation (SAO) parameter boundary of the current image unit is shifted to the left by xs lines from the vertical boundary of the current image unit, and the current image unit’s Horizontal Sample Offset Compensation (SAO) parameter boundary shifts ys rows upwards from the horizontal boundary of the current image unit.Step 1240, according to one or more spatial loop filter parameters, the limit of the spatial loop filter of the current image unit The pixels processed by Sample Offset Compensation (SAO) above the shrinkage boundary are subjected to spatial loop filter processing, wherein the bottom edge of the current image unit is shifted upward by yV lines to obtain the shrinkage of the spatial loop filter of the current image unit Boundary. In order to reduce line buffer requirements and/or reduce loop filter processing switching, loop filter related boundaries are set according to positive integers m, xs, ys, and yv, where xs is greater than m, ys is greater than or equal to 0, and ys is less than yv, and yv is set according to m (as shown in step 1250).

The flowchart shown above is intended to illustrate an example of loop filtering processing according to the present invention. Those skilled in the art can modify the steps, rearrange the order of the steps, divide the steps, or combine the steps to implement the present invention without departing from the spirit of the present invention. In this disclosure, specific syntax and semantics have been used to describe the embodiments. Those skilled in the art can implement the present invention by substituting equivalent syntax and semantics without departing from the spirit of the present invention.

The above description enables those skilled in the art to practice the present invention in accordance with the disclosed context and conditions of specific applications. Those skilled in the art may change the content of the above embodiments in various ways, and the general principles defined in this specification may be applied in other embodiments. Therefore, the present invention is not intended to be limited to the particular embodiments shown and described above, but is to be accorded the widest scope consistent with the principles and novel features described. In the above detailed description, descriptions of various specific details are used to help a thorough understanding of this case. Those skilled in the art will understand that the present invention can be practiced.

The embodiments of the present invention as described above can be realized by various hardware, software codes, or a combination of both. For example, an embodiment of the present invention may be that one or more electronic circuits are integrated into a video compression chip, or program codes are integrated into video compression software to execute the processes described herein. An embodiment of the present invention may also be program code, executed by a digital signal processor (DSP), to perform the processing routines described herein. This case may also involve a number of functional blocks implemented by computer processors, digital signal processors, microprocessors, or field-effect programmable logic arrays (FPGAs). These processors may be configured to perform specific tasks according to the present invention, by executing machine-readable software code, or by executing firmware code that defines specific methods embodied by the invention. Software code or firmware code may be developed in different programming languages and in different formats or styles. The software code is also compiled for different target platforms. However, different coding formats, styles, and software code languages and other coding means for performing tasks according to the present technology will not depart from the spirit and scope of the present invention.

The present invention may be embodied in other specific forms without departing from the spirit or essential characteristics of the inventions. The described examples are considered to be illustrative in all respects only and not restrictive. Accordingly, the scope of the invention is indicated by the claims rather than the foregoing description. All changes within the methods and ranges equivalent to the claims belong to the scope of the present invention.

Claims (26)

1. a kind of method of loop filtering processing, it is characterised in that methods described is entered for video coding system Row rebuilds video data, and above-mentioned reconstruction video data is divided into multiple image units, and methods described includes：

Receive the reconstruction video data of image unit；

To rebuild pixel implement block elimination filtering processing, wherein, block elimination filtering processing two image units it Between the corresponding horizontal end in image unit border each side correct up to m pixel；

According to one or more sample value migration parameters, to being handled in current image unit through block elimination filtering The pixel crossed carries out the sample value migration parameter side of sample value migration processing, wherein current image unit All or part of pixel in boundary shares same said one or multiple sample value migration parameters, wherein The vertical sample value migration bound of parameter of current image unit by current image unit vertical boundary to the left Displacement x s rows, and current image unit horizontal sample value migration bound of parameter by current image unit Horizontal boundary shifts up ys rows；And

According to one or more space loop filter parameters to hollow loop filter of current image unit Limit the pixel that sample value migration has been treated above border and make space loop filtering process, wherein when The space loop filter of preceding image unit limits border and shifts up yv from the base of current image unit OK；

Wherein, m, xs, ys and yv are positive integer, and xs is more than m, and ys is more than or equal to 0, and ys Less than yv, and yv is set according to m.

2. the method for loop filtering processing according to claim 1, it is characterised in that each shadow As unit one code tree unit of correspondence.

3. the method for loop filtering processing according to claim 1, it is characterised in that above-mentioned space Loop filtering processing correspondence adaptive loop filter processing.

4. the method for loop filtering processing according to claim 1, it is characterised in that above-mentioned reconstruction Video data is included at luma component and chrominance components, above-mentioned block elimination filtering processing, sample value migration Reason and space loop filtering process point luma component and chrominance components are carried out, the above-mentioned m each used Value is marked as M and N, and the above-mentioned xs each used is marked as xS and xSC, the above-mentioned ys each used Marked as yS and ySC, and the above-mentioned yv each used is marked as yV and yVC.

5. the method method of loop filtering according to claim 4 processing, it is characterised in that yS with YSC is equal to 0.

6. the method for loop filtering processing according to claim 4, it is characterised in that yV is more than M and yVC are more than N.

7. the method for loop filtering processing according to claim 6, it is characterised in that yV is equal to (M+1) and yVC be equal to (N+1).

8. the method for loop filtering processing according to claim 7, it is characterised in that M is equal to 3 And N is equal to 2.

9. the method for loop filtering processing according to claim 4, it is characterised in that yS is equal to YSC, yV are equal to yVC, and yVC is more than MAX (M, N).

10. the method for loop filtering processing according to claim 9, it is characterised in that yV and yVC Equal to MAX (M, N)+1.

11. the method for loop filtering processing according to claim 10, it is characterised in that M is equal to 3 and N is equal to 2.

12. the method for loop filtering processing according to claim 1, it is characterised in that further By the current pixel of the current line handled in the currently processed stage of current image unit and subsequent processing rank Symbol data obtained by the comparison of the adjacent pixel of the adjacent lines handled in section is stored, above-mentioned symbolic number According to corresponding to " being more than ", " being less than " or " being equal to ".

13. the method for loop filtering processing according to claim 12, it is characterised in that above-mentioned symbol Number is stored with two bits.

14. a kind of device of loop filtering processing, it is characterised in that described device is entered for video coding system Row rebuilds video data, and above-mentioned reconstruction video data is divided into multiple code tree units, and described device bag Include：

One or more electronic circuits, are coupled to line buffer, and be configured as：

Receive the reconstruction video data of image unit；

To rebuild pixel implement block elimination filtering processing, wherein, block elimination filtering processing two image units it Between the corresponding horizontal end in image unit border each side correct up to m pixel；

According to one or more sample value migration parameters, to being handled in current image unit through block elimination filtering The pixel crossed carries out the sample value migration parameter side of sample value migration processing, wherein current image unit All or part of pixel in boundary shares same said one or multiple sample value migration parameters, wherein The vertical sample value migration bound of parameter of current image unit by current image unit vertical boundary to the left Displacement x s rows, and current image unit horizontal sample value migration bound of parameter by current image unit Horizontal boundary shifts up ys rows；And

According to one or more space loop filter parameters to hollow loop filter of current image unit Limit the pixel that sample value migration has been treated above border and make space loop filtering process, wherein when The space loop filter of preceding image unit limits border and shifts up yv from the base of current image unit OK；

Wherein, m, xs, ys and yv are positive integer, and xs is more than m, and ys is more than or equal to 0 and small In yv, and yv is set according to m.

15. the device of loop filtering processing according to claim 14, it is characterised in that each Image unit one code tree unit of correspondence.

16. the device of loop filtering processing according to claim 14, it is characterised in that above-mentioned sky Between loop filtering processing correspondence adaptive loop filter processing.

17. the device of loop filtering processing according to claim 14, it is characterised in that above-mentioned heavy Building video data report includes luma component and chrominance components, and above-mentioned block elimination filtering processing, sample value skew are mended Repay processing and space loop filtering process point luma component and chrominance components are carried out, what is each used is upper M values are stated marked as M and N, the above-mentioned xs each used is marked as xS and xSC, and what is each used is upper Ys is stated marked as yS and ySC, and the above-mentioned yv each used is marked as yV and yVC.

18. the device of loop filtering according to claim 17 processing, it is characterised in that yS with YSC is equal to 0.

19. the device of loop filtering processing according to claim 17, it is characterised in that yV is big It is more than N in M and yVC.

20. the device of loop filtering processing according to claim 19, it is characterised in that yV etc. It is equal to (N+1) in (M+1) and yVC.

21. the device of loop filtering processing according to claim 20, it is characterised in that M is equal to 3 and N is equal to 2.

22. the device of loop filtering processing according to claim 17, it is characterised in that yS is equal to YSC, yV are equal to yVC and yVC is more than MAX (M, N).

23. the device of loop filtering according to claim 22 processing, it is characterised in that yV with YVC is equal to MAX (M, N)+1.

24. the device of loop filtering processing according to claim 23, it is characterised in that M is equal to 3 and N is equal to 2.

25. the device of loop filtering processing according to claim 14, it is characterised in that current shadow As unit the currently processed stage in the current pixel of current line that handles with being handled in subsequent processing stage Adjacent lines adjacent pixel comparison obtained by symbol data be stored, above-mentioned symbol data correspondence " More than ", " be less than " or " be equal to ".

26. the device of loop filtering processing according to claim 25, it is characterised in that above-mentioned symbol Number is stored with two bits.