H.265解码器去块滤波并行化设计与性能优化  

H.265 Decoder Deblocking Filter Parallel Design and Performance Optimization

作　　者：周建政 刘华平 

机构地区：[1]天格科技(杭州)有限公司,浙江杭州310011 [2]上海格谱信息科技有限公司,上海200072

出　　处：《电视技术》2015年第14期13-16,共4页Video Engineering

基　　金：杭州市重大科技创新项目(20142011A07)

摘　　要：H.265继续沿用H.264编码架构,去方块滤波器也是H.265视频编码标准的一个重要选项,去除混合编码带来的块效应极大改善了视频的质量,但由于H.265超级宏块的存在,去方块效应滤波相关参数层层嵌入在每个小的处理单元中,这种结构不利于实现基于宏块行间的并行化,同时也很难高效地利用Cortex-A9架构SIMD优化性能。首先详细分析H.265标准去块滤波器的处理过程以及并行处理的困难,进而提出一种便于实现基于宏块行间的并行去块滤波结构,然后进行Cortex-A9汇编优化。基于HM14.0实验,改进去方块效应滤波器计算复杂度从占整个解码器25%降至14%,大大提升了解码器性能,为移动设备上实现H.265大分辨率视频实时播放奠定基础。H.265 continue to use the H.264 encoding framework of the algorithm, which an important option tool is deblocking filter, the tool remove the block effect of hybrid coding and greatly improved the video quality, but because of the existence of H.265 super macro block, deblocking filter parameters embedded in a processing unit for each small, this structure cannot achieve parallel base on macro block lines, meanwhile it is also very difficult to effective use of Cortex-A9 architecture for SIMD performance optimization. This paper has a detailed analysis of the H.265 standard deblocking filter and the difficulty of parallel processing, and then proposed a convenient implementation of parallel base on macro block lines, and then assembly optimization in the new structure. The proposed deblocking filter algorithm is impledmented on HM14.0. Compared with the existing deblocking filter structure in HM14.0, the computational complexity decreased from 25% to 14%, greatly enhance the performance of the decoder, make sure H.265 high resolution video can real-time player on mobile devices.

关 键 词：去块滤波 H.265 并行化 解码器 

分 类 号：TN949.6[电子电信—信号与信息处理]