完全可伸缩视频编码的实现和改进  

Achievement and improvement for full-scalable video coding

作　　者：向友君[1] 张克新[1] 吴宗泽[1] 谢胜利[1] 

机构地区：[1]华南理工大学电子与信息学院,广州510640

出　　处：《计算机应用研究》2009年第9期3578-3580,共3页Application Research of Computers

基　　金：国家杰出青年科学基金资助项目(60325310);国家自然科学基金重点资助项目(U0635001)

摘　　要：提出了一种完全可伸缩视频编码的改进方法。首先给出了一种采用运动补偿时域滤波、二维离散小波变换和EZW编码的传统可伸缩编码方案,该方案将时间、空间、质量三方面的伸缩性有机地结合在一起,实现了完全可伸缩性能。针对运动补偿时域滤波技术中的GOP(group of pictures)结构和更新操作进行了改进,引入GOP结构自适应选择,降低系统内存需求和运算复杂度;采用基于人眼视觉系统特性的最小可觉差对更新操作中所引进的误差进行内容自适应修正。实验结果表明,编码效率和视频序列的重建质量都得到了显著的提高。This paper presented an improved method for full-scalable video coding. First, offered a full-scalable video coding system, which was based on the motion compensated temporal filter （MCTF）, discrete wavelet transforms （DWT） and embed- ded zero tree wavelet（EZW）. In this system, organized the coded bit-stream to achieve the brilliant combination of three main scalabilities： temporal, spatial and PSNR scalabilities. Then, presented adaptive GOP structure and content adaptive update steps based on the property of the human vision system for lifting-based motion-compensated temporal filtering （MCTF）. Ex- perimental results show that the coding efficiency and the quality of reconstructed sequence are improved significantly.

关 键 词：可伸缩视频编码 运动补偿时域滤波 图像组 最小可觉差 

分 类 号：TN919.8[电子电信—通信与信息系统]