基于DCT域的任意分辨率视频可伸缩编码

Scalable Video Coding of Arbitrary Resolution in DCT Domain

空间分辨率可伸缩编码能够满足网络视频对不同码率的要求以及不同用户显示终端的需求,成为近几年研究热点.由此提出一种基于DCT域准卷积的任意分辨率下采样方法,该方法通过分析子块与宏块DCT系数之间的关系,实现了由子块DCT系数预测宏块DCT系数,通过低通滤波器截取预测的宏块低频DCT系数,实现任意分辨率下采样.实验证明:该方法适合诸如H.264、AVS等基于DCT变换的视频编码标准,且提供灵活的空间可伸缩能力.

Spatial scalable video coding technique becomes a hot problem in recent years because this kind of multi-resolution coding technique can satisfy with the requirement of online video transmitting in low bit-rate and various terminal devices displaying.In this paper,we propose a scalable video coding of arbitrary resolution in DCT domain based on convolution.Firstly,we estimate each macro-block′s coefficients based on its sub-block coefficients referring to the relationship between the sub-block coefficients and the macro-block coefficients.Secondly,in order to reduce the original video resolution,a filter is designed to get the low frequency components of the estimated macro-block.Finally,these low frequency components can be saved or transmitted as one part of the current video coding stream,and a low computing complexity and an arbitrary on-demanded resolution video stream can be obtained,it is own to different size of low-pass filter and a scheme which can get macro-block DCT coefficients directly.Experimental results demonstrate that our scheme can provide flexible spatial scalable ability,and it can be used to process all kinds of the DCT-based video encoders including AVS and H.264,etc.