CN100574460C - AVS inter-frame predicated reference sample extraction method - Google Patents

Abstract

The invention provides a kind of AVS inter-frame predicated reference sample extraction method, described method is predicted the macro block of M * N in the picture frame, each macro block comprises four sub-pieces 0～3 of brightness and two sub-pieces 4 of colourity and 5, and described method makes the reference pixel of extraction be stored in the RS respectively: row reference memory, row back-up registers, zero point back-up registers, zero point backup of memory and row back-up registers.The present invention has not only saved the memory logic resource; And improved storage efficiency, can satisfy the requirement of real-time high definition coding.

Description

AVS Intra Prediction Reference Sample Extraction Method

technical field

The invention relates to a digital video encoding technology, in particular to an AVS intra-frame prediction reference sample extraction method.

Background technique

AVS (Audio Video coding Standard) is a digital audio and video codec technology standard independently developed by my country. Intra-frame prediction is one of the technical means to improve coding efficiency. It uses the adjacent pixels of the current block to directly predict each coefficient. The correlation between adjacent blocks is removed more effectively, thereby greatly improving the efficiency of intra-frame coding. AVS video standard 9.4.2 defines the intra prediction mode, 9.8 defines the intra prediction formula and the definition of reference pixels, AVS intra prediction includes luma intra prediction and chrominance intra prediction, a total of 33 reference samples are required, 16 1 upper reference sample, 16 left reference samples, 1 zero reference sample. The intra-frame prediction data volume and calculation amount are very large, and it is one of the bottlenecks in the operation of the entire system. To improve the coding performance and meet the requirements of real-time high-definition coding is a key problem to be solved by the intra-frame coding technology.

After searching the existing literature, it was found that in the article "AVS Intra-frame Prediction Algorithm and Its Hardware Implementation of Decoder" published by Wang Zheng and Liu Peilin in the journal "Computer Engineering and Application" in 2006, a possible Reconstruct the design of the computing unit, but the article does not give a reference sample extraction method.

Contents of the invention

The technical problem solved by the present invention is to provide a method for extracting AVS intra-frame prediction reference samples. The method reasonably arranges the storage space during sample extraction, reduces the consumption of logic resources to a certain extent, and can achieve real-time high-definition Coding requirements, and facilitate the realization of hardware.

In order to solve the above technical problems, the present invention adopts the following technical means: an AVS intra-frame prediction reference sample extraction method, the method predicts macroblocks in M rows and N columns in an image frame, and each macroblock contains 0 to 3 Four luminance sub-blocks and two chrominance sub-blocks of 4 and 5, the storage space is composed of a row reference memory, a row backup register, a zero point backup register, a zero point backup memory and a column backup register, wherein:

The row reference memory includes memory A for storing the bottom row of pixels of luma sub-blocks 2 and 3; and memory B for storing the bottom row of pixels of chrominance sub-blocks 4 and 5;

The line backup register, denoted as register F, is used to save the bottom row pixels of the luminance sub-blocks 0 and 1, and the pixels are used to provide reference pixels for the macroblock luminance sub-blocks 2 and 3;

The zero point backup register, denoted as register E, is used to save the zero point reference pixels of brightness sub-blocks 0 and 1;

The zero-point backup memory includes a memory C for storing the zero-point reference pixels of the luma sub-blocks 2 and 3; and a memory D for storing the zero-point reference pixels of the chrominance sub-blocks 4 and 5;

Column backup registers, including register G, which holds the rightmost column pixels of luma subblocks 0 and 1; register H, which holds the rightmost column pixels of luma subblocks 2 and 3; and register I, which holds the chrominance The rightmost column pixels of sub-blocks 4 and 5;

Said method comprises the following steps:

Step 1, extract the row reference samples of all sub-blocks 0, 1, 4 and 5 of the first row of macroblocks, whose values are all 0; extract the columns of all sub-blocks 0, 2, 4 and 5 of the first row of macroblocks Reference sample, its value is also 0; this step does not store;

Step 2, extract the zero-point reference samples of all luminance sub-blocks 0 and 1 of the first row of macroblocks, as the value of register G, extract the zero-point reference samples of all chrominance sub-blocks 4 and 5 of the first row of macroblocks, as registers The value of I; extract the zero point reference samples of all brightness sub-blocks 0 of the first row of macroblocks, as the value of memory A, extract the reference samples of all brightness sub-blocks 2 of the first row of macroblocks, as the value of register F, extract Reference samples of all chrominance sub-blocks 4 and 5 of the first column of macroblocks, as the value of memory B;

Step 3, extract the row reference samples of sub-blocks 0 and 1 of all macroblocks in a frame of image, as the value of memory A, extract the reference samples of all sub-blocks 2 and 3, and extract all sub-blocks 4 as the value of register F and 5 row reference samples as the value of memory B;

Step 4, extract the column reference samples of sub-blocks 0 and 1 of all macroblocks in a frame image, as the value of register G, extract the column reference samples of all sub-blocks 2 and 3, and extract all sub-blocks as the value of register H Reference samples of 4 and 5, as the value of register I;

Step 5, extract the zero-point reference samples of sub-blocks 0 and 1 of all macroblocks in a frame image, and extract all sub-blocks 2 and 3 zero-point reference samples as the value of memory C, and extract all sub-blocks as the value of register E Zero-point reference samples of 4 and 5, as the value of memory D.

Since the present invention adopts the above intra-frame prediction reference sample extraction method, the extracted reference pixels are stored in the register and the memory respectively, which saves memory logic resources; the reference sample extraction of each sub-block only needs 2 to 3 clock cycles, The efficiency is improved, and the requirement of real-time high-definition coding can be met.

Description of drawings

The method for extracting reference samples for intra-frame prediction suitable for AVS of the present invention is given in detail by the following embodiments and accompanying drawings.

Fig. 1 is a schematic diagram of arrangement of macroblocks and sub-blocks thereof according to the present invention;

Detailed ways

The method for extracting reference samples for intra-frame prediction of the present invention will be further described in detail below.

Figure 1 is a schematic diagram of the arrangement of M rows and N columns of macroblocks and their sub-blocks. In an embodiment of the present invention, it is assumed that M=N=2 in the image frame shown in FIG. A macroblock with two rows and two columns, each macroblock contains four luminance subblocks of 0 to 3 and two chrominance subblocks of 4 and 5, and reference samples useful for predicting other subblocks have been marked with digital grids 1 to 8 , in this embodiment, different storage spaces are arranged for reference samples of different sub-blocks, and the storage space is composed of a row reference memory, a row backup register, a zero point backup register, a zero point backup memory and a column backup register, wherein:

The row reference memory includes memory A for storing the bottom row of pixels of luma sub-blocks 2 and 3; and memory B for storing the bottom row of pixels of chrominance sub-blocks 4 and 5;

The line backup register, denoted as register F, is used to save the bottom row pixels of the luminance sub-blocks 0 and 1, and the pixels are used to provide reference pixels for the macroblock luminance sub-blocks 2 and 3;

The zero point backup register, denoted as register E, is used to save the zero point reference pixels of brightness sub-blocks 0 and 1;

The zero-point backup memory includes a memory C for storing the zero-point reference pixels of the luma sub-blocks 2 and 3; and a memory D for storing the zero-point reference pixels of the chrominance sub-blocks 4 and 5;

Column backup registers, including register G, which holds the rightmost column pixels of luma subblocks 0 and 1; register H, which holds the rightmost column pixels of luma subblocks 2 and 3; and register I, which holds the chrominance The rightmost column pixels of sub-blocks 4 and 5;

Said method comprises the following steps:

Step 1, extract the row reference samples of all sub-blocks 0, 1, 4 and 5 of the first row of macroblocks, whose values are all 0; extract the columns of all sub-blocks 0, 2, 4 and 5 of the first row of macroblocks Reference sample, its value is also 0; this step does not store;

Step 2, extract the zero-point reference samples of all luminance sub-blocks 0 and 1 of the first row of macroblocks, as the value of register G, extract the zero-point reference samples of all chrominance sub-blocks 4 and 5 of the first row of macroblocks, as registers The value of I; extract the zero point reference samples of all brightness sub-blocks 0 of the first row of macroblocks, as the value of memory A, extract the reference samples of all brightness sub-blocks 2 of the first row of macroblocks, as the value of register F, extract Reference samples of all chrominance sub-blocks 4 and 5 of the first column of macroblocks, as the value of memory B;

Step 3, extract the row reference samples of sub-blocks 0 and 1 of all macroblocks in a frame of image, as the value of memory A, extract the reference samples of all sub-blocks 2 and 3, and extract all sub-blocks 4 as the value of register F and 5 row reference samples as the value of memory B;

Step 4, extract the column reference samples of sub-blocks 0 and 1 of all macroblocks in a frame image, as the value of register G, extract the column reference samples of all sub-blocks 2 and 3, and extract all sub-blocks as the value of register H Reference samples of 4 and 5, as the value of register I;

Step 5, extract the zero-point reference samples of sub-blocks 0 and 1 of all macroblocks in a frame image, and extract all sub-blocks 2 and 3 zero-point reference samples as the value of memory C, and extract all sub-blocks as the value of register E Zero-point reference samples of 4 and 5, as the value of memory D.

The extraction of sub-block reference samples in all macroblocks in a frame of image is completed, and the reference sample extraction of each sub-block only needs 2 to 3 clock cycles, requiring less logic resources and high efficiency.

Claims (1)

1, a kind of AVS inter-frame predicated reference sample extraction method, described method is predicted the macro block of M * N in the picture frame, each macro block comprises four sub-pieces 0～3 of brightness and two sub-pieces 4 of colourity and 5, it is characterized in that, described method is arranged different memory spaces to the reference sample of the sub-piece of difference, described memory space by row reference memory, row back-up registers, zero point back-up registers, zero point backup of memory and row back-up registers form, wherein:

The row reference memory comprises memory A, is used to preserve the end row pixel of the sub-piece 2 of brightness and 3; And memory B, be used to preserve the end row pixel of the sub-piece 4 of colourity and 5;

The row back-up registers is designated as register F, is used to preserve the end row pixel of the sub-piece 0 of brightness and 1, and described pixel is used to the sub-piece 2 and 3 of this macro block brightness that reference pixel is provided;

Zero point, back-up registers was designated as register E, was used to preserve the reference pixel at zero point of the sub-piece 0 of brightness and 1;

Zero point, backup of memory comprised memory C, was used to preserve the reference pixel at zero point of the sub-piece 2 of brightness and 3; And memory D, be used to preserve the reference pixel at zero point of the sub-piece 4 of colourity and 5;

The row back-up registers comprises register G, is used to preserve the right column pixel of the sub-piece 0 of brightness and 1; Register H is used to preserve the right column pixel of the sub-piece 2 of brightness and 3; And register I, be used to preserve the right column pixel of the sub-piece 4 of colourity and 5;

Described method comprises following steps:

Step 1, all sub-pieces 0,1,4 that extract the first row macro block and 5 capable reference sample, its value all is 0; Extract all sub-pieces 0,2,4 of the first row macro block and 5 row reference sample, its value also all is 0; This step is not stored;

Step 2, the sub-piece 0 of all brightness that extracts the first row macro block and 1 reference sample at zero point as the value of register G, extract the sub-piece 4 of all colourities of the first row macro block and 5 reference sample at zero point, as the value of register I; Extract the reference sample at zero point of the sub-piece 0 of all brightness of the first row macro block, as the value of memory A, extract the reference sample of the sub-piece 2 of all brightness of the first row macro block, as the value of register F, extract the sub-piece 4 of all colourities of the first row macro block and 5 reference sample, as the value of memory B;

Step 3, the sub-piece 0 that extracts all macro blocks in the two field picture and 1 capable reference sample as the value of memory A, extract the reference sample of all sub-pieces 2 and 3, as the value of register F, extract the capable reference sample of all sub-pieces 4 and 5, as the value of memory B;

Step 4, the sub-piece 0 that extracts all macro blocks in the two field picture and 1 row reference sample as the value of register G, extract the row reference sample of all sub-pieces 2 and 3, as the value of register H, extract the reference sample of all sub-pieces 4 and 5, as the value of register I;

Step 5, the sub-piece 0 that extracts all macro blocks in the two field picture and 1 reference sample at zero point, as the value of memory C, extract the reference sample at zero point of all sub-pieces 2 and 3, as the value of register E, extract the reference sample at zero point of all sub-pieces 4 and 5, as the value of memory D.