Research On Fast Intra Prediction Algorithm In H.264/AVC Based On Texture Analysis

H.264/AVC is the newest video-coding standard published by the international standards bodies ITU-T (International Telecommunication Union) and ISO/IEC (International Organization for Standardization / International Electro-technical Commission). Intra prediction is an important constituting part of H.264/AVC, which includes many prediction modes. On prediction, it predicted the macro-block using all prediction modes by calculating RDO, and selected the prediction mode with least RDO cost as the best prediction mode. The variety of prediction mode and high computation complexity of RDO cost lead that the intra prediction algorithm is the one of time consuming procedure. In this paper, through the analysis of the original algorithm and studying the intra-prediction modes and algorithms, a new fast intra prediction algorithm based on textual analysis is proposed.The main work in this thesis covers the following aspects:1,This paper introduces the gray-tone spatial-dependence matrix and its features which determines the macro-block flat level and textural property. According to the flat level of macro-block, it will be divided into flat and non-flat macro-block which is used to determine that Intra₄×4 or Intral 6×16 is available for prediction, and to the textural property, it will be divided into the overall flat macro-block and a macro-block with texture direction which is used to exclude the impossible mode and reduce the range of candidate prediction mode. In the procedure of determining the textual property of macro-block with the inverse difference moment feature of the gray-tone spatial-dependence matrix, to strengthen the difference of macro-block with different textual property, improve the algorithm of inverse difference moment and achieve desirable effects.2,To reduce the complexity of 4×4 block intra prediction , the proposed algorithm researched the correlation of best prediction mode of 4×4 block in one 8×8 block. Firstly, select many standard video sequences whose macro-block mostly adopt intra₄×4 prediction mode, and got best prediction mode of all 4×4 block by full-search algorithm based the reference program JM13.2. Then, analyze the best prediction mode of all 4×4 block in one 8×8 block. Experiment results show that the best prediction mode of four 4×4 blocks has great correlation. Base on this conclusion, we can determine the candidate prediction mode of four 4×4 blocks by the textural features of corresponding 8×8 block. This algorithm will reduce the number of candidate prediction mode and candidate prediction mode of all four 4×4 blocks will be made sure at one time. The proposed algorithm greatly reduced the complexity of intra prediction algorithm.3,Base on the reference program JM13.2, proposed algorithm is realized. Experiment results show that the proposed fast algorithm can achieve remarkable computation reduction while maintaining similar PSNR and bit rate performance contrast with full search algorithm.