H.264宏块选择的自适应算法研究

A study on algorithm of adaptive block division and motion estimation in H.264

在视频编码的过程中,运动估计占据了举足轻重的地位.其性能优劣会在一定程度上决定了码流质量.采用16×16宏块能够提高压缩比,但容易产生方块效应,使用4×4宏块可以提高图像质量,但降低了压缩比。H.264视频压缩标准采用多模式运动估计,可以有效减少块匹配预测误差,但随着模式的增多,算法计算量成倍增加.为了克服这个困难,本文提出一种新的自适应的宏块划分和运动估计算法。这种分块算法在综合考虑图像本身特性在编码过程中,根据图像的各个部分运动程度不同而采用不同的分快策略,兼顾图像质量和处理开销。

Motion estimation plays an important role in video encoding. Its performance affects the quality of encoded bit stream greatly. Using 16×16 macro block leads to serious mosaic affection, Although it may be help to enhance the compression ratio. Using 4×4 macro block can improve picture quality easily comparing with previous method, but it sacrifice the high compression ratio. Through the using of multi block pattern, H.264 gets the balance between quality and compression ratio. With the increasing of macro block patterns, computing complexity increases sharply. In order to avoid this dilemma, this paper provides a new algorithm to accomplish this goal. According to the moving character of the picture, the algorithm adopts diverse policy to get the balance between quality and complexity.