Analysis And Real-time Implementation Of Fast Mode Decision Algorithm For HEVC

People’s need for high definition video is surging with improving technology, pursuiting higher quality of videos with lower bitrate. HEVC, the High Efficiency Video Coding standard, can achieve better performance compared with H.264/AVC. It can reduce bit-rate by 50% when remaining the same video quality. At the same time, its coding complexity is greatly improved than H.264/AVC. And Ultra high definition video encoder struggles with high complexity as well as huge data amount and real-time problems compared with high definition video encoder. The study of HEVC fast mode decision algorithm and its ultra high definition real-time implementation have application value.The fast mode decision algorithm of HEVC enhances the velocity of encoder, and parallel computing brought by multi-processor makes solution possible for real-time ultra high definition encoder implementation. In this paper, we begin with introducing a PU depth selection fast algorithm for UHD and HD HEVC encoder based on statistic distribution of variance for HEVC intra prediction. Then we design and implemente a HEVC UHD real-time encoder system step by step from algorithm analysis and data decomposition direction based on Tile-Gx36 platform and encoder modul characters. Futhermore, performance of HEVC UHD real-time encoder based on different encoder modul is analysesd in order to get better RD-curves. Based on the test result, we optimize the HEVC encoder system, improving the performance of HEVC encoder system largly and making a HEVC ultra high definition quasi real-time encoder system.In general, this paper provides the implementation of ultra high definition quasi real-time encoder system based on low power consumption multi-processor CPU. We not only realize our goal, but also provide platform foundation and theoretical basis for future work.