基于小波空间的可伸缩视频编码

文章提出了一种新的下采样技术提高了可伸缩视频编码的性能。可伸缩性是从空间、时间、SNR(质量)三个方面来提供的,同时还保持了高压缩率。其中基本层视频序列是从上一层经过下采样得到的视频序列。一次下采样后,分辨率降为原视频序列的1/2,基本层的运动向量与增强层的运动向量有很大的相关性。目前的做法是使用M PEG滤波器来进行下采样,层间的运动向量的关系减弱了。文章提出使用小波变换对输入的增强层的视频序列进行下采样,基本层就是小波空间的低频部分,这样在基本层的运动估计补偿可以通过相位平移在低频子带中预测,减少运动估计残差,同时,增强层的运动向量可以由基本层小波空间中的运动向量给出估计值。

Scalable Video Coding via the Wavelet Domain

This paper presents an advanced down-sampling technique for improved performance in Scalable Video Coding(SVC).SVC offers temporal,spatial,fine-granularity SNR scalability at the same time,while it still maintains high compression efficiency.For spatial scalability the video sequence of lower layer is obtained with down-sampling the video sequence of higher layer and its resolution is 1/2 resolution of higher layer after down-sampling filter.There are more correlations between the lower layer of motion information and the higher layer of motion information.The MPEG down-sampling filter reduced the correlations.So these correlations should be exploited carefully for excellent compression efficiency.We proposed a new down-sampling method which is implemented with Discrete Wavelet Transform(DWT).Motion-Compensated Prediction(MCP) is performed in subband-to-subband through phase-shifting matrix.Motion infor-mation of enhancement layer is derived from motion information of base layer.