基于模式复制的H.264多描述视频编码  

Mode Duplication-Based Multiple Description Coding

作　　者：董萌[1] 蔡灿辉[1] 

机构地区：[1]华侨大学信息科学与工程学院,厦门361021

出　　处：《信号处理》2011年第11期1675-1679,共5页Journal of Signal Processing

基　　金：国家自然科学基金(60772164);福建省自然科学基金(2011J01354)

摘　　要：本文提出了一种新的基于H.264的多描述视频编码算法——基于模式复制的多描述编码算法。首先对输入视频序列中的每一帧图像分别进行水平方向下采样和垂直方向下采样,形成四个子图像。相应的子图像构成四个视频子序列。把这四个子序列两两组合,形成两个描述,每个描述包含两个子序列。由于每个描述中两个子序列之间具有很强的空间相关性和时间相关性,其对应宏块的最佳模式和运动矢量基本相同,因此只需用H.264编码器对其中一个子序列进行编码,另一子序列则可直接采用上述已编码子序列的最佳模式和运动向量对其进行预测编码。这样只需要对其中一个子序列进行模式选择,也只需要对一个子序列的最佳模式和运动向量进行编码传输,既降低了计算复杂度,又提高了编码效率。实验结果表明,在中高码率下,本文算法与同类算法在相同比特率情况下,PSNR有明显的提高,并且比特率越高,这种优势就越明显。This paper presents a novel multiple description video coding algorithm for H.264,called mode duplication based multiple description coding.Each frame in the input video sequence is down-sampled first horizontally and then vertically to form four sub-frames.The resulted four sub-sequences are pair-wisely grouped to form two descriptions.Considering that two sub-sequences in a description have strong spatial correlation and temporal correlation,so the best modes and motion vectors in corresponding macro-blocks are basically the same.In this paper,only one sub-sequence per description is coded by a H.264 coder,and the other sub-sequence is coded by using the best modes and motion vectors of the aforementioned encoded subsequence.Consequently,only one sub-sequence per description needs to perform mode decision,reducing the computational complexity and bit rates.The experimental results have shown that at moderate and high rates,the proposed algorithm achieves a higher coding quality compared with other H.264 based MDC algorithms.

关 键 词：H.264/AVC 多描述编码 空间下采样 模式复制 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]