Design And Implementation Of Parallel Algorithms For Next Generation Video Coding Techniques

In today’s booming information age, digital video is concerned due to the efficient and ituitive information expression. With the puting out of H.264 and HEVC video coding standard, high-performance video compression has become realistic, but the computational complexity of codec greatly improves, severely restricting the speed of video codec.In recent years, GPU technology develops rapidly. Its floating point caculation and parallel computing ability become more and more suitable for general-purpose computing. In 2008, the proposing of computing framework makes designing parallel speedup algorithms based on heterogeneous platform consisted by CPU and GPU become posssible. Therefore, this paper deigns and implements several parallel algorithms for H.264 decoder and CABAC coding in HEVC with the use of GPU many-core environment, so as to improve the speed of HD video codec and meet the needs of actual application.This paper mainly includes three parts. Firstly, design algorithms to high-complexity modules in H.264 decoder. It includes designing total parallel algorithms to inverse quantization and IDCT, designing different-granularity parallel algorithms to luma 1/4-pixel interpolation and chroma 1/8-pixel interpolation, and designing diagonal parallel and total parallel algorithms separately to intra recovery and inter recovery. Secondly, design parallel algorithm to CABAC module in HEVC which spends too much running time, and mainly design two different-granularity total parallel algorithms to syntax elements preprocessor. Thirdly, design CPU dual-thread+GPU mutiple-thread parallel scheme to the whole H.264 decoder and HEVC coder, so as to enhance the codec acceleration effect.All parts above are implemented on CPU+GPU heterogeneous platform through OpenCL programming. Large numbers of experiments have been done in this paper, which indicates that comparing with the standard JM13.2 decoder, H.264 parallel decoder designed in this paper greatly improves the decoding speed, meanwhile ensuring the same decoding performance, and achieves the application requirements of HD video real-time decoding and reconstruction(at least 30 fps); Comparing with x265, CABAC parallel algorithm of HEVC in this paper improves the computational efficiency, and reduces the running time significantly.