摘要
针对人眼对图像不同区域的关注度存在的差异,本文提出了基于人眼视觉关注度的率失真优化帧内宏块刷新算法,该算法采用端到端的全局率失真理论,对图像中关注度不同的区域采用不同的处理方式,使高关注度区域的宏块能得到更及时的帧内刷新。由于受到码率的限制,必须对不同区域采用不同的码率分配机制。最后,在JVT的JM12.2平台上实现该算法。大量实验结果表明,与传统的基于率失真优化的帧内刷新算法相比,基于视觉关注度的率失真优化帧内刷新算法在高关注度区域能够取得更好的主观质量,并且在保证关注度区域主观质量的同时整个码率有所降低,因此非常适合无线网络上的视频传输。
The importance of video sequence varied with the regions, and human visual attention too varied with the importance of region of video sequence, so the subjective quality requirement of human visual attention high region is not the same as human visual attention low region. The paper proposed the RDO Intra Refreshment algo- rithm based on human visual attention, and it used the end to end overall distortion theory to analysis human visual attention high region and human visual attention low region, and would given more IR chances to the MB in human visual attention high region, as the bit rate was constrained, so the bit rate of human visual attention high region and human visual attention low region, It must be reallocated. At last, the algorithm was implemented in reference software JM 12.2. Simulation results showed that, comparing with traditional algorithm, the proposed could obtain better subjective quality of reconstructed video in visual attention region.
基金
国家自然科学基金项目(60772106)
湖北省教育厅优秀中青年人才计划项目(Q20082803)
