Research And Optimization On H.264/AVC Intra Mode Fast Selection Algorithm

H.264/AVC standard was established together by the experts of ISO/IEC MPEG(Moving Picture Experts Group) and ITU-T VCEG (Video Coding Experts Group). As the latest video coding standard, H.264/AVC standard introduces a large number of advanced technologies of the previous audio and video coding international standards, and adopts a number of new technologies to improve the coding effect. That makes H.264/AVC acquire much higher coding efficiency and better network interactive. But importing the new technologies also brings great computational complexity for the encoder of H.264/AVC, which limits the application of H.264/AVC in various fields.Firstly, this paper briefly introduces the principle and commonly-used technology of video coding, and then it generalizes the current digital video coding standard such as MPEG-1, MPEG-2 and H.263, etc. Next, the paper makes a detail introduction of H.264/AVC standard, summarizing the history of its development, the encoder and the decoder's structure and the main technical feature of each module. Finally, this paper makes a further study on the H.264/AVC intra prediction technology, and summarizing the process of best encoding mode of the luminance blocks and chrominance blocks. Because of the diversity of intra prediction mode and complexity of rate-distortion cost function, the process of intra prediction becomes one of the more time-consuming processes. Based on the analysis of the H.264/AVC standard intra prediction process, this paper proposes a fast selection algorithm about the best mode of intra prediction.The major job of my paper focuses on the following several aspects:(1) This paper introduces the concept of image distortion fuzzy metric criteria, which was proposed by Lai, and improves the fuzzy metric criteria. Then we apply the fuzzy metric criteria in the H.264/AVC intra prediction process. Thus the intra mode selection process of H.264/AVC standard is speeded up.(2) The concept of macroblock fuzzy metric is proposed, which is used to fast determine the flat degree of macroblock. The best encoding mode of detail-rich macroblock can be directly identified as intra₄×4 mode, thus we can avoid the cost calculation of different prediction modes of intra₁6×16, which accelerates the encoding speed.(3) Statistics found that intra₄×4 mode covers a large proportion in the whole intra mode (85% above). Therefore, this paper adopts the square gradient to determine the texture direction of the 4×4 block in advance, and then eliminates the mode which is inconsistent with the texture direction. So it can speeds up the process of intra₄×4 mode selection, thus reducing the computational amount of the whole encoder.(4) Due to the calculation principle of image distortion fuzzy metric and the complexity of the calculation of square gradient, and also considering the process unit of H.264/AVC is 16×16 macroblock, this paper introduces the relevant fast algorithm. Some coefficients can be set as constants to further reduce the computational amount, which accelerates the computational speed.In order to verify the functions of the proposed fast algorithm, this paper adopts the reference model JM13.2 of H.264/AVC as an experimental platform and implements the fast algorithm. The experimental results are compared with full search algorithm results. The results show that under the premise of an average reduction of 0.06dB in PSNR and encoding bit rate increased by 1.62%, the fast algorithm reduces 52.14% of the encoding time, which significantly improves the encoding speed.