分布式视频编码中量化失真和LDPC编解码失真的优化

针对分布式视频编码中的量化失真和LDPC（Low Density Parity Check）编解码失真问题,提出了基于边缘分类的DCT（Discrete Cosine Transform）系数重组和基于边缘Hash的子块修复方案.首先根据子块边缘方向的不同,WZ （Wyner-Ziv）帧子块被分成3类,对不同类型的子块,其DCT系数采用不同的扫描方式并重组,使较大的DCT系数被组合在一起,减小了交流系数的量化步长.同时,把子块边缘信息作为Hash信息,指导WZ帧LDPC解码出错的子块的修复.最后,通过实验证实了提出的方法能使高量化等级下编码器的PSNR（Peak Signal to Noise Ratio）性能增益达到0.5 dB左右.

Quantization Distortion and LDPC Codec Distortion Optimization for Distributed Video Coding

Quantization distortion and Low Density Parity Check （LDPC） codec distortion exist in distributed video coding. Response to these problems an edge classification based Discrete Cosine Transform （DCT） coefficients reorganization and edge Hash based block repair scheme is proposed. Blocks in Wyner-Ziv （WZ） frame are first classified into 3 types according to edge directions. Then different scanning order is applied to the DCT coefficients of each class of blocks, and the coefficients are reorganized. In this way, larger coefficients may be grouped together and quantization step for Alternating Current （AC） coefficients may be reduced. Meanwhile, the edge information of the block is taken as Hash information to assist in block repair for blocks with LDPC decoding error in WZ frame. Finally, experiments proved that the proposed scheme might make performance gain of the encoder improve about 0.5 dB at high quantization levels.