最优扫描下新预测模式的帧内预测算法  

An intra prediction algorithm for new prediction modes under best scanning

作　　者：李宏伟[1] 吴成柯[1] 宋锐[1] 

机构地区：[1]西安电子科技大学ISN国家重点实验室,西安710071

出　　处：《高技术通讯》2011年第2期162-166,共5页Chinese High Technology Letters

基　　金：国家自然科学基金（60802076）,高等学校创新引智计划（B08038）,长江学者和创新团队发展计划（PCSITR）资助项目.

摘　　要：提出了一种最优扫描下新预测模式的帧内预测算法。首先利用边缘相关性选取最优的扫描方式，然后通过边缘方向估计和统计概率的联合预处理方法缩减单向预测模式的数量，并通过加权内插模型和啥达马变换值（SATD）确定双向预测模式，最后采用最优扫描方式下所选的预测模式进行编码。仿真结果表明，与双扫描双参考帧内预测（DSM_BIP）算法相比，本文算法使输出码率降低了2．84％，重建图像的平均峰值信噪比（PSNR）提高了0．12dB，而且严格地控制了编码时间。An intra prediction algorithm for new prediction modes at best scanning is proposed. The proposed algorithm determines the best scanning modes first according to the edge relativity, then uses edge orientation estimation and statistical probability to decrease the number of unidirection prediction modes, and determines the bi-direction intra prediction mode through the weighted interpolation value and the sum of absolute transformed differences （STAD）. Finally the selected prediction modes under the best scanning are encoded. The simulation results show that, compared with the double scannings and bi-direction intra prediction （DSM-BIP） algorithm, the peak signal noise ratio （PSNR） is increased by 0.12dB and the bitrate is reduced by 2.84% when using the proposed algorithm, moreover, the coding time is restricted dramatically.

关 键 词：最优扫描 预测模式 帧内预测 边缘相关性 边缘方向估计 

分 类 号：TP334.22[自动化与计算机技术—计算机系统结构]