MPEG—2视频编码算法的几点改进

在ＭＰＥＧ—２ＴＭＳ［１］的基础上，本文提出如下几项改进：１）根据视觉特性，引入亮度掩盖算子修正ＴＭ５的自适应量化级计算方法，节省了码字而不影响主观质量。２）为了均匀分配亮度掩盖节省的码字，以当前Ｐ、Ｂ帧中具有较大均方差值的帧内量化宏块为编码难度的测度在编码过程中自适应地调整预置的帧码字数。这一措施也是处理场景切换的一种可行方式。３）在运动估值中综合考虑ＭＡＤ值和运动矢量编码占用比特，进一步提高编码效率，并使运动矢量场的一致性有所改善。计算机模拟实验结果表明，采用本文方法，平均视频解码信噪比提高了０．３ｄＢ以上。由于在视觉敏感区平均量化级有明显下降，图像的主观质量得到改善。

Some Improvements on MPEG-2 Video Coding Algorithm

Compared with MPEG-2 TM5,there are three changes proposed in this paper: i)Based on the visual perception, a brightness masking operator is introduced to modify the TM5 adaptive algorithm of calculating quantization scales. More bits are saved by the modification without percepual degradation. ii)To uniformly distribute the bits saved in i), the target bit allocation of P and B frame can be adaptively changed during encoding by a difficulty measure which depends upon the existence of large variance intra frame quantization macroblock in current frame. The method is also a possible way to handle the scene changes. iii) By a simple combined minimization of the MAD value and coding bits of motion vector, the efficiency of coding and the consistence of motion vectors are improved slightly. The computer simulation experiment results show that the average decoding SNR is increased more than 0. 3 dB by using these improvements. Because of the obvious reduction of the average quantization values in visual sensitive regions, the decoding image quality is percetually improved.