基于宏块相关性的帧间预测模式选择策略  

Rate-control Algorithm of H.264 /AVC Based on Human Visual System

作　　者：牛丽君[1] 宋娜[1] 张刚[1] 王园园[1] 

机构地区：[1]太原理工大学信息工程学院,山西太原030024

出　　处：《电视技术》2014年第5期28-30,共3页Video Engineering

基　　金：国家自然科学基金项目(60772101)

摘　　要：视频编码中,相邻帧之间的景物存在一定相关性,因此各宏块帧间预测模式存在一定相关性。在AVS标准中,帧间预测模式选择是单纯地从代价值大小的角度进行判断,忽略了当前帧与参考帧之间的相关性以及人眼视觉的感官性。为使宏块的预测模式更加合理、更能反映主观质量,提出了基于宏块相关性的帧间预测模式选择策略。在大量实验基础上得出阈值κ来调节率失真代价值与宏块相关性间的平衡关系。综合考虑代价值与阈值κ得出当前块的帧间预测模式。实验结果表明:该算法不仅能提高信噪比PSNR与结构相似度SSIM,而且能使码率略微降低。In video coding, a certain correlation exists in the inter prediction mode of macro blocks, attributed to the fact that there is a correlation be- tween adjacent frames. In the AVS standard, the selection of inter prediction mode purely depends on its cost, which ignores the correlation between cur-rent frame and reference frame as well as the human senses. To make the inter prediction mode more reasonable and reflect the subjective quality further, a new selection strategy based on correlation of macro blocks is proposed in this paper. On the basis of massive experiment, the balance relationship be- tween the rate-distortion cost and the correlation of macro blocks adjusted by the threshold K is concluded in this paper. The best inter prediction mode of the current block is acquired considering rate-distortion cost and the correlation between macro blocks comprehensively. The experimental results indi-cate that,this algorithm can not only reduce bit rate, but also improve PSNR and SSIM.

关 键 词：宏块相关性 帧间预测 AVS 

分 类 号：TN919.8[电子电信—通信与信息系统]