一种新的用于屏幕图像编码的HEVC帧内模式

摘要

由于传统编码方式对屏幕图像的编码效果不佳，该文根据屏幕图像包含大量非连续色调内容的特点，在HEVC(High Efficiency Video Coding)基础上，提出一种新的帧内编码模式称为帧内串匹配(Intra String Copy, ISC)。基本思想是在HEVC的编码单元(Coding Unit, CU)级别上，引入字典编码工具：编码时，在一定长度的字典窗口内，利用散列表，对当前CU内的像素，进行串搜索和匹配；解码时，根据像素串匹配的距离和匹配长度，在重建缓存内复制相应位置像素重建当前CU像素。实验结果表明，在编码复杂度增加很少的情况下，对于典型的屏幕图像测试序列，在全帧内(All Intra, AI)，随机接入(Random Access, RA)，低延迟(Low-delay B, LB)3种配置下，有损编码模式比HEVC分别节省码率15.1%,12.0%,8.3%，无损编码模式分别节省码率23.3%,14.9%,11.6%。%Because of the poor effect of the traditional coding methods on the screen content coding, considering the screen content is rich in non-continuous tone content, a new intra coding mode based on High Efficiency Video Coding (HEVC), which is called Intra String Copy (ISC), is proposed. The basic idea is adopting the dictionary coding tool on the HEVC Coding Unit (CU) level. When encoding, the current CU pixels are searched and matched in a certain length dictionary window by using Hash table. When decoding, according to the pixels string matching distances and lengths, the current CU pixels in the reconstruction cache are restored by copying the corresponding position pixels. Experiment results show that with little coding complexity increase than HEVC, for typical screen content test sequences, ISC can achieve lossy coding bit-rate saving of 15.1%, 12.0%, 8.3% for All Intra (AI), Random Access (RA), and Low-delay B (LB) configurations, respectively, and lossless coding bit-rate saving of 23.3%, 14.9%, 11.6% for AI, RA, and LB configurations.