Research On Key Technologiesof H.264/HEVC Video Coding & Decoding Standard

As the multimedia application more and more widely spreads, it puts forward higher and higher requirements on the image resolution, frame rate, real-time performance, which brings a big challenge to the video coding&decoding standard. Besides, for some other applications, such as space communications, deep space exploration, etc., they always have limited bandwidth, higher BER(bits error ratio), and are vulnerable to the environmental impact. If we want to transmit the image within the limited bandwidth and ensure good quality in the meantime, the video compression technology needs to be improved urgently. Based on the above background, this paper studied the H.264 and HEVC video coding&decoding standards, and made some optimizations and improvements, which improved the encoding&decoding efficiency. What’s more, the H.264/HEVC high definition video coding&decoding platforms were also built, and their real-time performance and parallel processing strategies were studied.The main work includes the following:After the deep research of the framework and key technologies of H.264, this paper mainly focuses on the inter motion estimation algorithm and has made some improvements. By analyzing the advantages and shortcomings of the current H.264 fast motion estimation algorithms, it proposes a new algorithm based on “local extremum search”. It divides the area to be searched into sub-blocks, then searches for the local extremum in each sub-block and gets the global extremum by comparing all the local extremums. It could make the search avoid being trapped in local extremums, which achieves very high matching accuracy. In the meanwhile, it takes some early termination and fast jump-over strategies, which speeds up the search process greatly. The experiment results show that the proposed algorithm improves the coding efficiency of H.264 and has goodapplicationvalue.In order to study the real-time applications of H.264, this paper builds a high definition video coding&decoding platform using the DM6467(ARM+DSP) microprocessor, and completes the development of the functionalmodules, such as high-definition camera driver, image capturing, automatic dimming control, DSP real-time video compressing, networkcommunication, decoding and displaying. Finally, it realizes the 1080p@30f H.264 video coding&decoding. Test results show that the platform has low output bitrate and good real-time performance, which meets the application requirements quite well.Later, this paper takesHEVC, the new generation video coding&decoding standard, as research object, mainly focusing on its inter motion estimation algorithm and high accuracy intra prediction algorithm for lossless coding.In the aspect of inter motion estimation, this paper combines the partial distortion search together with the traditional fast search algorithms, proposing a new algorithm based on the “multi-traps assumption”. By dividing the PU block into sub-blocks, computing the image complexity of each sub-block, estimating the number of traps, using the sub-block to do the coarse search, refining the final search, etc., it could achieve high matching accuracy and reduce the searching complexity in the meantime. Results show the proposed algorithm is very suitable for the real-time inter motion estimation, especially for the high resolution videos.In the aspect of intra prediction, considering the fact that the lossless compression performance of HEVC is inferior to that of H.264/AVC and even the MPEG, it needs to be improved to make the HEVC more outstanding andadaptable to different applications. This paper mainly focuses on two typical intra prediction algorithms: the prediction based on the “DPCM transform in the residual domain” and the “pixel-based angular prediction”. By analyzing their advantages and disadvantages, some improved algorithms are proposed: the prediction based on the “multi-transforms in the residual domain” and the “pixel-based adaptive directional prediction”. Experimental results show that the proposed algorithms outperform others in both compression gain and encoding&decoding speed, which demonstrate their superiority. As the lossless video coding becomes more and more important in some applications, these researches could play important roles in the future.In the last, in order to study the relationship between the huge computational complexity and the real-time performance of HEVC, this paperbuilds the 1080p@60f real-time video coding &decoding platform using the DSPC-8681(with 4 C6678 multi-core DSPs on it). It also tests the coding performance of HEVC when there are different DSP cores participating in the coding process. Based on the analysis of the results, this paper discusses the prospects of HEVC.