近似像素合并优化的HEVC屏幕视频编码算法

Optimization of Merging Approximate Pixels for HEVC Screen Content Coding

由于计算机生成的屏幕视频和摄像机拍摄的自然视频在本质上有很大差异,计算机屏幕视频具有特殊的图像特征,使用传统视频压缩工具对其进行压缩时,编码效果很不理想.提出一种算法,对屏幕视频中出现的近似像素进行合并处理,有效结合了基于串拷贝机制的字典编码方案,并应用于HEVC视频编码标准平台.近似像素合并的过程中,综合考虑了像素出现频度和像素差异等因素,在不损伤图像主观质量的前提下,该算法能明显改善屏幕视频编码方案的性能.实验结果表明,在全帧内(All Intra,AI)和低延迟(Low-delay B,LB)两种配置下,和ISC(Intra String Copy)算法比较,所提算法的码率平均节省了1.4%和0.35%.和HEVC屏幕视频编码测试模型SCM4.0比较,码率平均节省了10.39%和5.5%.

Since computer screen content have its distinct characteristics compared with the natural camera captured video content,traditional video compression tools are not suitable for screen contents.In this paper,we propose a novel algorithm for screen content coding.After merging the approximate pixels in screen image,the proposed method efficiently combined with the dictionary coding scheme which based on string copy mechanism,and applied to High Efficiency Video Coding reference software.Pixels history occurrence frequency and pixels difference are considered when we design the approximate pixel merging algorithm,the proposed method can obviously improve the performance of screen content coding scheme without harming the subjective image quality of screen video.Experimental result shows that the proposed method can achieve Bitrate saving of 1.4% and 0.35% against ISC（Intra String Copy） method for All-Intra （AI） and Low-delay B （LB） configuration respectively.In comparison with the original HEVC SCC reference software SCM4.0,Average bitrate saving of 10.39% and 5.5% can be achieved for AI and LB configuration respectively.