标题:
分布式压缩视频感知综述A Survey on Distributed Compressed Video Sensing
作者:
解晨, 龚声蓉
关键字:
：分布式视频编码, 压缩感知, 字典学习, 稀疏表示, 信号重构Distributed Video Coding; Compressive Sensing; Dictionary Learning; Sparse Representation; Signal Reconstruction
期刊名称:
《Journal of Image and Signal Processing》, Vol.2 No.1, 2013-01-30
摘要:
本分布式视频编码(Distributed video coding, DVC)是一种全新的视频编解码技术。与传统的联合编码、联合解码不同，分布式视频编码对两个或多个独立同分布的信源进行独立编码，然后由单一解码器利用信源之间的相关性对所有编码的信源进行联合解码，故分布式视频编码使得低复杂度编码成为可能，从而很好地解决了视频编码终端受限的情况。另一方面，压缩感知(Compressive Sensing, CS)为信号采样提供了新的方式，它基于信号的稀疏性、测量矩阵的随机性和非线性优化算法对信号的压缩测量重构，从而突破了传统Nyquist采样定理的限制。压缩感知理论应用于分布式视频编码，能够使编码端更加低复杂化、低消耗、简单化。本文主要综述了分布式视频编码相关理论、当前的经典方案，以及压缩感知理论和现阶段分布式压缩视频感知的发展现状、涉及到关键方法。最后分析了现存方案的一些问题和思路，并讨论了其未来可能的发展方向。
This electronic distributed video coding is a new paradigm for video compression. Compared to conventional video coding standards in which the video sequence is coded jointly and decoded jointly, distributed video coding sys-tem codes the video sequence separately for two or more sources that are independent identically distributed and de-codes jointly with the statistical correlation between different sources, then the coder becomes as simple as possible, so as to solve the problems of the limited video terminal. On the other hand, an emerging signal acquisition technology (Compressive Sensing, CS) provides a new way for the signal sampling, signal compression reconstruction based on the sparstiy of signal, random measurement matrix and nonlinear optimization algorithm. It broke through the limitations of traditional Nyquist sampling theorem, which has been applicable to directly capture compressed image data efficiently. Combination of distributed video coding and CS (Distributed Compressed Video Sensing, DCVS) results in more low-complexity and low-cost for video coding. This paper reviews the theory of distributed video coding, classic schemes involved, as well as theoretical knowledge of compressed sensing, and development status of distributed com-pressed video sensing at this stage. Finally we present some problems and the probably corresponding solutions, then discuss its possible applications in the future prospects.