Abstract: Compressed Video Sensing（CVS） is a video encoding method, which combines Compressed Sensing（CS） with Distributed Video Coding（DVC）. It is also called Distributed Video Compressed Sensing. In CVS, each frame of video is subjected to block partitioning, compression sampling, and then DPCM is performed on the data. Finally, quantization is performed using uniform or non-uniform quantization. At present, the design of the CVS quantizer is mostly designed on the premise that the sampled data or residual data obey the Gaussian distribution. This paper analyzes the distribution features of the data after image block division, compressed sampling, and DPCM by Kolmogorov-Smirnov test. On the basis of those analysis, Lloyd’s optimal quantizer design criteria is used in the proposed quantizer. Experimental results show that proposed quantizer has reduced about 14.2% in BD-Rate and improved about 0.11?dB in BDPSNR compared to the traditional quantizing method, improves the compression efficiency and reconstruction quality of CVS.

Key words: compressed sensing, optimal quantizer, Compressed Video Sensing（CVS）, Differential Pulse Code Modulation（DPCM）

摘要： 压缩视频感知（Compressed Video Sensing，CVS）是一种利用压缩感知（Compressed Sensing，CS）以及分布式视频编码（DVC）的视频压缩方法，故又被称为分布式视频压缩感知。在CVS中，每帧图像经过块划分、压缩采样后对数据进行DPCM，最后使用均匀或者非均匀量化进行量化。目前，CVS量化器的设计大多是在采样数据或残差数据服从高斯分布的前提下设计的，通过Kolmogorov-Smirnov检验进一步分析压缩采样后的数据，利用劳埃德最佳量化器准则训练量化码书，设计出一种简单、高效的量化器。经实验，设计的量化器相比于传统的量化方法在BD-Rate上减少了约14.2%，在BDPSNR上提升了约0.11?dB，提高了CVS的压缩效率和重建质量。

关键词: 压缩感知, 最佳量化, 压缩视频感知, 差分脉码调制