Low complexity surveillance Video Coding based on Distributed Compressive Video Sensing

Abstract

In video surveillance applications, the foreground moving objects in the frame are segmented from its background and are coded with fewer bits compared to frame-based coding. In such techniques, encoder becomes complex due to object segmentation and object motion estimation. Actual motivation of compressive sensing on video is to have a simple encoder, and therefore, we propose Distributed Compressive Video Sensing based Video Object Compression (DCVS-VOC) technique in which, (i) the object segmentation is done only for certain frames, and (ii) the motion estimation is performed at the decoder instead of encoder. Experimental results show that the proposed DCVS-VOC is capable of handling any CS reconstruction algorithm at its decoder.