CN101247525A - A Method of Improving the Intra-Frame Coding Rate of Image - Google Patents

Abstract

The present invention discloses a method for increasing the encoding speed in the image frame basing on the H.264 standard. The method filters the estimating mode along the grain direction according to the grain direction of the image, and then executes RDO calculation thereby greatly reducing the calculating time of the RDO. The method of the invention effectively increases the image encoding speed on the base that the excellent image quality is kept.

Description

A Method of Improving the Intra-Frame Coding Rate of Image

technical field

The invention relates to video compression coding technology, in particular to a method for improving the intra-frame coding rate based on H.264 images.

Background technique

There are many methods of video/image coding, and its applications involve all aspects. In order to enable terminals produced by different manufacturers to exchange information with each other, or receive information from a common signal source, some international organizations began to work on video in the late 1980s. /Standardization of image coding. At the same time, major manufacturers are interested in this, and directly promote the research process of video/image coding standardization. Finally, in 1988, the International Telephone and Telegraph Consultative Committee (CCITT, International Telephone and Telegraph Consultative Committee, renamed ITU-T, International Telecommunication Union Telecommunications Standardization Sector in 1992) formulated the first video coding standard - H.261, thus becoming a video Milestones in coding history. Subsequently, ISO MPEG (Moving Picture Experts Group) and ITU-T VCEG (Video Coding Experts Group) were integrated according to different application environments and requirements, and based on H.261 and the core, a series of video coding standards were successively formulated. ITU-T focuses on the field of real-time video communication applications, and released the H.26x series of standards (such as H.261, H.262, H.263 and H.264, etc.), ISO MPEG is mainly for video storage media, TV broadcasting and For multimedia communication applications, MPEGx series standards (such as MPEG-1, MPEG-2 and MPEG-4, etc.) have been formulated. Figure 1 shows the development process of video coding standards. The following briefly introduces the development process of international video coding standards.

In 1985, CCITT organized the 15th expert group to further study the standardization of conference TV, and in 1988, it formulated the H.261 standard for 64kbit/s video telephony/conference application ^[30] . In the H.261 standard, intra-frame (I frame, Intra frame) coding and inter-frame (P frame, Predictive frame) coding are defined, and technologies such as inter-frame prediction, DCT transformation, and Huffman coding are used. In order to enhance flexibility, H.261 only makes strict restrictions on code stream syntax, code stream multiplexing, decoding process, etc. There are no restrictive regulations on parts such as controls that have an important impact on reconstructed image quality indicators but do not affect compatibility, providing developers, manufacturers and users with a large space for activities. The successful introduction of the H.261 standard has greatly encouraged major manufacturers, international standard organizations and scientific research institutes, and set off a climax of research and application of video compression coding.

In 1991, ISO MPEG began to formulate the MPEG-1 standard. The main goal is to establish a standard for moving images, related sounds and their combination coding suitable for digital storage, media storage and retrieval, and it became an international standard in November 1993. The MPEG-1 standard takes H.261 as the basic frame, introduces bidirectional prediction frame (B frame, Bi-directional prediction frame), half-pixel precision motion estimation, and the concept of picture group (GOP, Group Of Picture), implements random read Fetch, fast forward and backward search and reverse playback and other functions.

Subsequently, ISO MPEG and ITU-TVCEG initiated the formulation of the MPEG-2 draft ^[32] (also known as H.262 in the ITU-T standard series), and it was determined to become a standard in 1994. MPEG-2 is based on MPEG-1 and has made important extensions. For interlaced conventional TV images, the methods of "coding by frame" and "coding by field" are specially set; a binary code stream structure is created: program stream and transport stream, and the operating environment of transport stream There may be serious errors, but the operating environment of the program stream rarely has errors; according to the complexity of the encoding technology, the concept of profile and level is introduced for the first time, which cleverly solves the exchangeable bit stream sex and internationality. In addition, the concept of scalability is added, which allows video signals of different quality levels or different temporal and spatial resolutions to be obtained from a coded data stream. Scalability includes space domain, signal-to-noise ratio, time domain, etc. MPEG-2 is a very successful standard and has been widely integrated into people's life and work.

With the development and popularization of network technology, network bandwidth has increasingly become a bottleneck that hinders people from using video/image. In order to alleviate this problem, ITU-T VCEG proposed the H.263 coding standard for low bit rate applications in 1995. It is based on H.261 and attracts effective and reasonable technologies in recommendations such as MPEG1/2, and provides four optional coding algorithms, namely unconstrained motion vector algorithm, syntax-based algorithm coding, advanced prediction method and PB frame algorithm to further improve coding efficiency. In addition, H.263 expands the image format and supports images in various formats including QCIF, Sub-QCIF, CIF, 4CIF and 16CIF. In addition, the standard does not limit the number of frames per second, so the maximum rate can be limited by reducing the number of frames per second. On the basis of the H.263 standard, ITU-TVCEG proposed two work plans: one is the so-called Short Term (Short Term) plan, which is to add some new function options on the basis of H.263, further improve compression efficiency and expand some function, while starting a so-called long-term (Long Term) plan to develop a new international standard for low-bit-rate video communications. In the short-term plan, new versions of H.263+ and H.263++ will be launched successively. The long-term plan was to produce a standard called H.263L, but it was renamed H.26L in 1998.

In 1998, ISO MPEG proposed the MPEG-4 standard, which integrated technologies and functions in the fields of digital TV, interactive graphics and the Internet, and further expanded and supplemented on the basis of H.263, MPEG-1 and MPEG-2. . The encoding rate of MPEG-4 covers a large range from as low as 5kbit/s to higher than 2Mbit/s, and proposes a completely different encoding concept from previous image encoding standards. It absorbs the object-based encoding method Its coding scheme is based on the object model of arbitrary shape, and it adds shape information to the description of the image compared with the traditional coding standard. MPEG-4 is no longer a standardized fixed algorithm, but to establish an expandable coding tool set, and various algorithms are constructed by the tool set. Without decoding, it supports content-based processing and code stream editing, supports artificial image/sound and natural image/sound synthesis, supports content-based random access, etc., and has different application environments. Better robustness, support content-based scalable coding. At the same time, ISO MPEG has also developed standards such as MPEG-7 and MPEG-21 to provide a standardized description for various multimedia information, as well as an efficient, transparent and interoperable media framework.

In order to further improve video coding efficiency, ISO and ITU-T formally established the Joint Video Team (JVT) in December 2001, and began to work on the standardization of H.264/MPEG-4part 10 (AVC, Advanced Video Coding) ( In this paper, it is uniformly referred to as H.264), which was formally determined as an international standard in May 2003. Based on H.26L, it extends the application from low bit rate to high bit rate. In addition to inheriting the advantages of previous standards, H.264 also introduces many new technologies, so that under the same decoding quality conditions, the coding efficiency is nearly 50% higher than that of H.263 and MPEG-4.

In addition, it is worth noting that the Ministry of Information Industry of my country began to formulate an audio and video coding standard (AVS, Audio Video coding Standard) with independent intellectual property rights in June 2002. The goal of AVS is to formulate common technical standards for digital audio and video codec, processing and presentation, and to provide efficient and economical codec technology for digital audio and video equipment and systems, mainly for HDTV, HD-DVD, wireless broadband multimedia communication, Internet broadband Streaming media and other major information industry applications.

Same as the previous video coding standards, the H.264 system also adopts the MC-DCT structure, that is, the hybrid structure of motion compensation and transform coding. Its coding technology framework is shown in Figure 2. H.264 coding is mainly composed of intra prediction, inter prediction (motion estimation and compensation), integer transformation, quantization and entropy coding.

H.264 adopts two encoding modes of intra frame (Intra) and inter frame (Inter). Supported video source formats include (YUV) 4:2:0, 4:2:2 and 4:4:4, both progressive and interlaced video sequences are supported, and for interlaced video frames, H.264 support The odd and even fields are encoded independently, and the method of encoding the odd and even fields together is also supported. For I (intra-coded frame) frame images, intra-frame coding is used; for P-frame (forward predictive frame) and B-frame (bidirectional predictive frame) images, inter-frame mode coding is used, but at the macroblock layer, also Intra mode encoding can be selected. Coding is performed in units of non-overlapping macro-blocks (MB, Macro-Block), and macro-blocks are generally defined as 16×16 pixel blocks.

For an I-frame image, first perform intra-frame prediction, then carry out integer transformation and quantization on the prediction residual signal (the difference between the original value and the predicted value), and then perform variable-length coding or arithmetic coding on the quantized coefficients to generate a compressed code stream, and at the same time The image is reconstructed through processes such as inverse transformation and inverse quantization, so as to be used as a reference for subsequent frame encoding. For P-frame images, first perform high-precision motion estimation and intra-frame prediction of multi-mode and multi-reference frames, and select inter-frame and intra-frame coding modes and corresponding block modes according to rate-distortion optimization (RDO, Rate-Distortion Optimization), Then the residual signal is transformed, quantized and entropy coded to generate a compressed code stream, and the image is reconstructed through inverse transformation and inverse quantization. For B-frame images, similar to P-frame images, bidirectional prediction technology is first used for motion estimation and intra-frame prediction of multi-mode and multi-reference frames, and the best coding mode is selected according to rate loss optimization, and then the residual signal is transformed and quantized and entropy encoding to generate a compressed code stream. In addition, H.264 also defines SI and SP frames.

In order to improve the network adaptability of coding, H.264 adopts a coding structure in which video coding layer (VCL, Video Coding Layer) and network abstraction layer (NAL, Network Abstraction Layer) are separated, as shown in Figure 3, VCL completes the video image High-efficiency compression, NAL is responsible for packaging and transmitting data in the appropriate way required by the network.

H.264 achieves higher video compression ratio, better image quality and good network adaptability. Therefore, H.264 has a wide range of applications, including videophone (fixed or mobile), real-time video conferencing system, Video surveillance system, Internet video transmission and multimedia information storage, etc. In order to meet the characteristics of wide application, H.264 only stipulates the bit stream, syntax elements and decoding process, and does not restrict the encoder, which makes the implementation of the H.264 encoder very flexible.

Intra prediction of H.264

In H.264, the 9 optional prediction mode sets for 4×4 blocks are: {mode 0, mode1, ..., mode 8}, that is, the vertical prediction direction is mode 0, and the horizontal prediction direction is mode 1. And so on. 4 optional prediction mode sets for 16×16 blocks: {mode 0, mode 1, ..., mode 3}, corresponding to vertical, horizontal, DC and planar prediction directions. The 8×8 chroma prediction mode is the same as the 16×16 prediction mode. In addition to the DC prediction mode, each prediction mode has a corresponding prediction direction and prediction weight. The DC prediction method uses the average value of adjacent boundary pixels for prediction, that is, all pixel values of the prediction block are equal to the average of adjacent boundary pixels. value.

The basic process of intra prediction of a macroblock is as follows:

(1) Judging the available macroblock information around the current macroblock, including whether the upper, left, and upper right macroblocks are available.

(2) Divide the luminance macroblock into 16 4×4 small blocks, adopt 9 different prediction modes, and calculate the rate-distortion cost of each 4×4 small block in turn, and the rate-distortion cost is defined by the Lagrangian function:

J(s,c,mode|QP,λ _mode )=SSD(s,c,mode|QP)+λ _mode R(s,c,mode|QP)

Wherein, s is the original pixel block signal, c is the reconstructed block signal, QP is the quantization parameter of the macroblock, λ _mode is the Lagrangian multiplier, λ _mode =0.85·2 ^QP/3 , R(s, c, mode|QP) is the coding rate under the corresponding mode and QP, including header information, prediction mode and the number of coded bits of all DCT coefficients. SSD(.) is the sum of squared errors between the 4×4 original pixel block and the reconstructed block:

$\mathrm{<span\; class="notranslate">\; SSD</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; mode</span>}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; QP</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 3</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}$

The selection of all 4×4 intra-frame prediction modes is completed in sequence, and the sum of the rate-distortion costs of all 4×4 blocks is obtained.

(3) Using the 16×16 block mode, select the best prediction mode from 4 different prediction modes. The method is to calculate the SATD (Sum of Absolute Transformed Difference) value in all modes, and select the mode with the minimum value as the best 16× 16 prediction modes. The SATD value is used as the rate-distortion (RD) cost of the 16×16 block, which is half of the absolute value sum of the residual value divided into 16 4×4 blocks and subjected to Hardamard transformation, as shown in the following formula, DiffT(i, j ) is the Hardamard transform coefficient.

$\mathrm{<span\; class="notranslate">\; SATD</span>}<span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; DiffT</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; j</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; /</span>\mathrm{<span\; class="notranslate">\; 2</span>}$

(4) Comparing (2) and (3) to obtain the rate-distortion cost, and select an optimal prediction mode as the coding mode of the macroblock. Chroma blocks are computed the same as 16x16 blocks.

There are a total of N8×(N4×16+N16) mode combinations for luma blocks and chrominance blocks in a macroblock ^[7] . N8, N4 and N16 represent the prediction of chrominance blocks, 4×4 and 16×16 luma blocks, respectively. number of patterns. That is to say, to obtain the optimal RDO mode for a macroblock, a total of 592 different RDO calculations need to be calculated. It can be seen that the computational complexity of intra-frame prediction mode selection is very high, which affects the encoding speed of H.264.

In order to reduce the complexity of intra prediction, many scholars have conducted extensive research. For example, Feng Pan in the literature (F.Pan, X.Lin. Fast Mode Decision for Intra Prediction. ISO/IEC JTC1/SC29/WG11 and ITU-T SG16 Q.6, JVT-G013.doc, JVT 7th Meeting Pattaya II, Thailand , 7-14, March 2003., hereinafter referred to as the Feng Pan method) proposed a fast mode selection algorithm for intra-frame prediction. The basic idea is to first use the Sobel operator for edge detection to obtain the edge direction of the object in the image block, and then according to the edge direction Determine the corresponding candidate prediction mode; Changsung Kim proposed a feature-based intra-prediction fast algorithm in the literature [C.Kim.Feature-based intra-prediction modedecision for h.264.IEEE., 2004, pp:769-772.] , using SAD and SATD two factors to characterize features, and through complex calculations and comparisons, the final coding mode is obtained. These methods improve the speed of intra-frame coding to a certain extent, but at the same time introduce complex calculations in the early stage, so there are limitations. Among them, the algorithm proposed by FengPan is recognized as the best among many algorithms and has been adopted by the JVT organization.

Contents of the invention

The purpose of the present invention is to increase the image coding rate on the basis of maintaining good image quality.

To achieve the object of the present invention, the idea of the present invention is to increase the coding rate by reducing the number of candidate modes in the intra prediction mode set. According to the texture direction of the image, the method of the invention screens out prediction modes along the texture direction, and then performs RDO calculation.

The texture direction of the above image is obtained by the following method:

First, expand the adjacent boundary pixels into two lines, and then define 4 texture directions according to the 9 prediction directions proposed in the standard: 0°, 45°, 90° and 135°;

Then calculate the average gray level difference in the texture direction of two adjacent rows of pixels, and record them as D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ respectively, and then take the minimum value among D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ The corresponding direction is taken as the texture direction of the edge, and let:

D _min =min(D ₀ , D ₄₅ , D ₉₀ , D ₁₃₅ )

Then take the direction corresponding to the minimum value D _min among D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ as the texture direction of the edge.

Specifically, the values of D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ can be calculated by the following formulas:

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 45</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 90</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}{<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 135</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$

Wherein, I(x, y) is the gray value of the pixel at the (x, y) coordinate, (x ₀ , y ₀ ) is the pixel coordinate of the upper left corner of the predicted block, and N is the size of the coding block. In the present invention, the value of N is preferably 4, 8 or 16. When the value of N is 8 or 16, the direction corresponding to the minimum value D _min among D ₀ , D ₄₅ and D ₉₀ is taken as the texture direction of the edge.

The prediction mode along the texture direction can be done as follows:

Let 9 kinds of prediction mode candidate sets of 4×4 blocks be F _4×4 ={mode 0, mode 1,...,mode 8}, and 4 kinds of prediction mode candidate sets of 16×16 blocks be F _16×16 = {mode 0, mode1, ..., mode3}, the four prediction mode candidate sets of 8 x 8 blocks are F _{8 x 8} = {mode 0, mode1, ..., mode 3}, which are filtered out by the following method Prediction mode along texture direction:

(1) Determine available macroblock information around the current macroblock, including top, left, and top right macroblocks;

(2) If both the upper and left macroblocks adopt 16×16 prediction mode, then judge the intra-frame complexity of the current macroblock, if the complexity of the current macroblock is less than the specified threshold, then jump to step (4); the current macroblock The complexity of a block can be calculated by the following formula:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; m</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

where M is the macroblock size.

(3) 4×4 block prediction mode selection:

The macroblock is divided into 4×4 blocks, and the selection of the prediction mode candidate set adopts the following strategy:

1) If D _min =D ₉₀ , candidate set F _4×4 ={mode 0, mode 7, mode 5, mode2};

2) If D _min =D ₀ , candidate set F _4×4 ={mode 1, mode 8, mode 6, mode2};

3) If D _min =D ₄₅ , candidate set F _4×4 ={mode 4, mode 5, mode 6, mode2};

4) If D _min =D ₁₃₅ , candidate set F _4×4 ={mode 3, mode 7, mode 8, mode2};

After the above calculations, the candidate prediction modes are reduced from the original 9 to 4. The encoder uses the prediction mode in the candidate set to perform RDO calculation, find the best mode, and calculate the rate-distortion cost sum of all blocks;

(4) 16×16 luma block and 8×8 chroma block prediction mode selection:

The macroblock is predicted by 16×16 blocks, and the selection of the prediction mode candidate set adopts the following strategies:

1) If D' _min =D ₉₀ , candidate set F _16×16 ={mode 0, mode 2};

2) If D' _min =D ₀ , candidate set F _16×16 ={mode 1, mode 2};

3) If D' _min =D ₄₅ , the formula set F _16×16 ={mode 3, mode 2};

The 8×8 chrominance block uses the same prediction mode as the 16×16 block. After the above calculations, the prediction modes are reduced from the original 4 to 2. The encoder uses the prediction mode in the candidate set to perform RDO calculation to obtain the best mode;

(5) Compare the rate-distortion cost of (3) and (4), and select the minimum cost mode as the final coding mode.

The above method can reduce the calculation of RDO, effectively improve the encoding speed of the image within the frame, and the image quality and bit rate change little.

Description of drawings

Figure 1 shows the development history of international standards;

Fig. 2 is a schematic diagram of the H.264 encoding framework;

Fig. 3 is a schematic diagram of H.264 layered design;

Fig. 4 is the first frame image of Foreman (QCIF) sequence;

Fig. 5 is the 69th macroblock (16 * 16 macroblock) among Fig. 4;

Fig. 6 is the 12th 4×4 small block of Fig. 5;

Fig. 7 is a schematic diagram of extending two rows of adjacent borders of 4×4 small blocks;

Figure 8 is the texture definition direction;

Figure 9 is a Foreman sequence PSNR comparison;

Figure 10 is a comparison of Stefan sequence PSNR;

Figure 11 is a comparison of Carphone sequence PSNR;

Figure 12 is a comparison of Tempete sequence PSNR.

Detailed ways

Further illustrate the present invention below in conjunction with accompanying drawing. It should be understood that the following examples are only used to illustrate the present invention, rather than as a limitation of the present invention, and any modifications or replacements without departing from the spirit and essence of the present invention fall within the scope of the present invention.

Example 1

1. Texture direction estimation

As we all know, natural images have strong spatial correlation, and the texture direction between adjacent macroblocks is also very similar, especially for 4×4 small blocks, the correlation is stronger. Figure 4 shows the first frame image of the Foreman (QCIF) sequence, Figure 5 shows the 69th macroblock in the image, and the white box in Figure 6 shows the 12th 4×4 small block of the 69th macroblock . It can be seen from the figure that the texture trends between adjacent macroblocks (or blocks) are very similar. The texture direction estimation is performed on the 4×4 luma, 16×16 luma and 8×8 chrominance blocks respectively in the following.

(1) 4×4 block texture direction estimation

First, the adjacent boundary pixels are extended into two rows, as shown in Figure 7. According to the 9 prediction directions proposed in the standard, 4 texture directions are defined: 0°, 45°, 90° and 135°, as shown in Figure 8.

Then calculate the average gray level difference in the texture direction between two adjacent rows of pixels, and record them as D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ . Methods as below:

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 45</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 90</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$

${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 135</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$

$\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; <figure-callout\; id="0"\; label="y"\; filenames="S2008101025176E00024.png,S2008101025176E00031.png"\; state="\{\{state\}\}">y</figure-callout></span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$

Among them, I(x, y) is the gray value of the pixel at (x, y) coordinates, (x ₀ , y ₀ ) is the pixel coordinates of the upper left corner of the prediction block, N is the size of the coding block, in 4×4 N=4 in the block. Then take the direction corresponding to the minimum value among D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ as the texture direction of the edge, and make:

D _min =min(D ₀ , D ₄₅ , D ₉₀ , D ₁₃₅ )

16×16 luma block and 8×8 chroma block texture direction estimation

For 16x16 luma blocks and 8x8 chroma blocks, there are only vertical, horizontal and planar prediction directions, plus DC prediction mode. Therefore only 3 texture directions need to be defined: 0°, 45° and 90°. According to the same method as the 4×4 block, the minimum D' _min and D _min are obtained respectively, and this direction is used as the texture direction of the edge.

Through the above calculation, the texture direction of the coding block is preliminarily determined.

2. Intra-frame mode selection algorithm

Let 9 kinds of prediction mode candidate sets of 4×4 blocks be F _4×4 ={mode 0, mode 1,...,mode 8}, and 4 kinds of prediction mode candidate sets of 16×16 blocks be F _16×16 = {mode 0, mode1, . . . , mode 4}, the four prediction mode candidate sets for 8×8 blocks are F _8×8 ={mode 0, mode1, . . . , mode 4}.

The algorithm is described as follows:

(1) Determine available macroblock information around the current macroblock, including top, left, and top right macroblocks;

(2) If the upper and left macroblocks both adopt 16×16 prediction mode, then calculate the intra-frame complexity of the current macroblock:

${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; m</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

where M is the macroblock size, equal to 16 in this paper. If X _I is smaller than a threshold T (obtained by experimental test, equal to 256 in this example), then directly skip to (4), and use 16×16 macroblocks for prediction.

(3) 4×4 block prediction mode selection:

The macroblock is divided into 4×4 blocks, and the selection of the prediction mode candidate set adopts the following strategy:

1) If D _min =D ₉₀ , candidate set F _4×4 ={mode 0, mode 7, mode 5, mode2};

2) If D _min =D ₀ , candidate set F _4×4 ={mode 1, mode 8, mode 6, mode2};

3) If D _min =D ₄₅ , candidate set F _4×4 ={mode 4, mode 5, mode 6, mode2};

4) If D _min =D ₁₃₅ , candidate set F _4×4 ={mode 3, mode 7, mode 8, mode2};

After the above calculations, the candidate prediction modes are reduced from the original 9 to 4. The encoder uses the prediction modes in the candidate set to perform RDO calculations to find the best mode and calculate the rate-distortion cost sum of all blocks.

(4) 16×16 luma block and 8×8 chroma block prediction mode selection:

The macroblock is predicted by 16×16 blocks, and the selection of the prediction mode candidate set adopts the following strategy:

1) If D' _min =D ₉₀ , candidate set F _16×16 ={mode 0, mode 2};

2) If D' _min =D ₀ , candidate set F _16×16 ={mode 1, mode 2};

3) If D' _min =D ₄₅ , the candidate formula set F _16×16 ={mode 3, mode 2}; the chrominance blocks are selected in the same way. After the above calculations, the prediction modes are reduced from the original 4 to 2. The encoder uses the prediction modes in the candidate set to perform RDO calculations to obtain the best mode.

(5) Compare the rate-distortion cost of (3) and (4), and select the minimum cost mode as the final coding mode.

Experimental results and analysis

In order to verify the effectiveness of the algorithm, the experiment uses the H.264 reference model JM7.5 (H.S.Malvar and A.Hallapuro.Low-complexity transform and quantization in H.264/AVC.IEEE Trans.CSVT., vol.13(7 ), pp: 598-602, 2003.) as a platform in which to implement the method. The experimental method is to use CAVLC entropy coding, the reference frame is 2, the search range is 32, the Hardmard transform is used, the rate distortion optimization (RDO) is used, the sequence structure adopts IPPP and all I formats, and the quantization parameter selection: 28 and 32, for multiple Standard video sequences were tested.

The experimental results of this method are compared with the calculation results of the full model algorithm and the calculation results of the Feng Pan algorithm. The full model algorithm is a method of exhaustive search and calculation of all prediction models. Since the running speed of the algorithm has a lot to do with the hardware platform, the above three algorithms are tested on the same hardware platform, and the ratio of the test results is compared, so that the interference of the hardware environment to the algorithm can be eliminated, that is, the interference of the test sequence The encoding rate change (B_CHG), encoding time change (T_CHG) and image quality change (PSNR_CHG) are compared, and the calculation method is as follows:

$\mathrm{<span\; class="notranslate">\; B</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; CHG</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; Bits</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; ours</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Bits</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; all</span>}}{\mathrm{<span\; class="notranslate">\; Bits</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; all</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; 100</span>}<span\; class="notranslate">\; \%</span>$

$\mathrm{<span\; class="notranslate">\; T</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; CHG</span>}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; Time</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; all</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; Time</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; ours</span>}}{\mathrm{<span\; class="notranslate">\; Time</span>}<span\; class="notranslate">\; \_</span>\mathrm{<span\; class="notranslate">\; all</span>}}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; 100</span>}<span\; class="notranslate">\; \%</span>$

PSNR_CHG = PSNR_ours - PSNR_all

Wherein Bits_ours and Bits_all are respectively the number of bits produced using the method of the present invention and the full-mode algorithm, Time_ours and Time_all are respectively the encoding time using the method of the present invention and the full-mode algorithm, and PSNR_ours and PSNR_all are respectively the number of bits using the method of the present invention and the full-mode algorithm Image Quality. The calculation method of Feng Pan algorithm is the same.

1. IPPP structure, encoding 120 frames, frame rate = 30f/s, GOP = 15, the encoding results are compared as follows:

Table 1 Experimental results of QCIF standard sequence (QP＝28)

Table 2 Experimental results of CIF standard sequence (QP＝32)

It can be seen from Table 1 and Table 2 that for the IPPP structure, this method will increase the encoding speed by 21.5% on average compared with the full-mode method, while the image quality will only decrease by 0.026dB, and the code rate will increase by about 0.593% on average. At the same time, it can be seen that compared with the algorithm in literature [7], the code rate increase and the image quality decrease of the two are comparable to the full-mode algorithm, but the average improvement of this method is 2%. The main reason is that the Sobel algorithm used in the Feng Pan algorithm to estimate the edge has a relatively large amount of calculation. Figure 9 and Figure 10 below are the brightness PSNR comparison charts of Foreman (QCIF) and Stefan (CIF) sequences respectively. The dashed line is the image quality under the full-mode algorithm, the solid line is the image quality under this method, and the quality of the P frame image The drop is due to the degradation of the I-frame image quality used for prediction.

2. Full I-frame structure, encoding 120 frames, frame rate = 30f/s, the encoding results are compared as follows:

Table 3 Experimental results of QCIF standard sequence (QP＝28)

Table 4 Experimental results of CIF standard sequence (QP＝32)

Tempete 3.87 38.26 -0.129 3.54 43.66 -0.125 Stefan 4.12 37.64 -0.098 4.77 39.48 -0.084 Mobile 4.36 33.63 -0.135 4.51 37.89 -0.139 average 4.07 38.18 -0.117 4.18 42.36 -0.120

It can be seen from Table 3 and Table 4 that for the full I-frame structure, the performance of this method is more obvious, and the encoding speed will be increased by 43% on average compared with the full-mode method, while the image quality will only decrease by 0.12dB, and the code rate will increase by an average of 4.0 %about. Similarly, this method is also better than the literature [7] algorithm, the code rate increase and the image quality decline of the two are almost the same, but this method has an average improvement of nearly 4%. Figure 11 and Figure 12 below are the brightness PSNR comparison diagrams of Carphone (QCIF) and Tempete (CIF) sequences, respectively. The dotted line is the image quality under the full-mode algorithm, and the solid line is the image quality under this method.

The above experimental results show that this method can effectively improve the intra-frame encoding speed of images, while the image quality and bit rate change little.

Claims (7)

1. A method for increasing the intra-frame coding rate of an image based on the H.264 standard. According to the texture direction of the image, the method screens out the prediction mode along the texture direction, and then performs RDO calculation. 2. The method according to claim 1, wherein the texture direction of the image is obtained by the following method: First, expand the adjacent boundary pixels into two lines, and then define 4 texture directions according to the 9 prediction directions proposed in the standard: 0°, 45°, 90° and 135°; Then calculate the average gray level difference in the texture direction of two adjacent rows of pixels, and record them as D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ respectively, and then take the minimum value among D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ The corresponding direction is taken as the texture direction of the edge, and let: D _min =min(D ₀ , D ₄₅ , D ₉₀ , D ₁₃₅ ) Then take the direction corresponding to the minimum value D _min among D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ as the texture direction of the edge. 3. The method according to claim 2, characterized in that the values of D ₀ , D ₄₅ , D ₉₀ and D ₁₃₅ are calculated by the following method: ${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$ ${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 45</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$ $\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$ ${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 90</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; N</span>}}\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>$ ${\mathrm{<span\; class="notranslate">\; D.</span>}}_{\mathrm{<span\; class="notranslate">\; 135</span>}}<span\; class="notranslate">\; =</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; \&times;</span>\mathrm{<span\; class="notranslate">\; N</span>}}<span\; class="notranslate">\; (</span>\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; 2</span>}\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; +</span>$ $\underset{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; )</span>$ Wherein, I(x, y) is the gray value of the pixel at the (x, y) coordinate, (x ₀ , y ₀ ) is the pixel coordinate of the upper left corner of the predicted block, and N is the size of the coding block. 4. The method according to claim 3, wherein the value of N is 4, 8 or 16. 5. The method according to claim 4, wherein when the value of N is 8 or 16, the direction corresponding to the minimum value D _min among D ₀ , D ₄₅ and D ₉₀ is taken as the texture direction of the edge. 6. The method according to any one of claims 1 to 5, characterized in that the nine prediction mode candidate sets for 4×4 blocks are F _4×4 ={mode 0,mode 1,...,mode 8}, the 4 kinds of prediction mode candidate sets for 16×16 blocks are F _16×16 ={mode 0, mode 1,...,mode3}, the 4 kinds of prediction mode candidate sets for 8×8 blocks are F _8×8 ＝{mode 0, mode 1,..., mode 3}, then filter out the prediction mode along the texture direction by the following method: (1) Determine available macroblock information around the current macroblock, including top, left, and top right macroblocks; (2) If both the upper and left macroblocks adopt the 16×16 prediction mode, then judge the intra-frame complexity of the current macroblock, if the complexity of the current macroblock is less than the specified threshold, then jump to step (4); (3) 4×4 block prediction mode selection: The macroblock is divided into 4×4 blocks, and the selection of the prediction mode candidate set adopts the following strategy: 1) If D _min =D ₉₀ , candidate set F _4×4 ={mode 0, mode 7, mode 5, mode2}; 2) If D _min =D ₀ , candidate set F _4×4 ={mode 1, mode 8, mode 6, mode2}; 3) If D _min =D ₄₅ , candidate set F _4×4 ={mode 4, mode 5, mode 6, mode2}; 4) If D _min =D ₁₃₅ , candidate set F _4×4 ={mode 3, mode 7, mode 8, mode2}; The encoder uses the prediction mode in the candidate set to perform RDO calculation, find the best mode, and calculate the rate-distortion cost sum of all blocks; (4) 16×16 luma block and 8×8 chroma block prediction mode selection: The macroblock is predicted by 16×16 blocks, and the selection of the prediction mode candidate set adopts the following strategies: 1) If D' _min =D ₉₀ , candidate set F _16×16 ={mode 0, mode 2}; 2) If D' _min =D ₀ , candidate set F _16×16 ={mode 1, mode 2}; 3) If D' _min =D ₄₅ , the formula set F _16×16 ={mode 3, mode 2}; The 8×8 chrominance block adopts the same prediction mode as the 16×16 block, and the encoder uses the prediction mode in the candidate set for RDO calculation to obtain the best mode; (5) Compare the rate-distortion cost of (3) and (4), and select the minimum cost mode as the final coding mode. 7. The method according to claim 6, wherein the complexity of the current macroblock is calculated by the following formula: ${\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; I</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; abs</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; *</span>\mathrm{<span\; class="notranslate">\; m</span>}}\underset{\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\underset{\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 0</span>}}{\overset{\mathrm{<span\; class="notranslate">\; m</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}\mathrm{<span\; class="notranslate">\; I</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$ where M is the macroblock size.

Images