Improved Distributed Compressive Video Sensing Based on HEVC Motion Estimation

Abstract

The Distributed Compressive Video Sensing (DCVS) system combines advantages of compressive sensing and distributed video coding to adapt to the limited-resource video sensing and transmission environment. To improve the reconstruction quality of non key frame which is also called CS frame , the reconstructed key (K) frames are used to generate side information (SI) frames of the CS frames. Therefore, the quality of SI frames greatly affects the reconstruction results of CS frames. However, in conventional distributed compressed video sensing schemes, the quality of SI frames does not achieve the ideal. Because of this, in the current study, a new motion estimation (ME) method called High Efficiency Video Coding motion estimation (HEVC-ME) is proposed for generating more accurate SI to improve the reconfiguration effect of CS frames. In the proposed HEVC-ME, a better estimation result is obtained by performing motion estimation with coding units (CU) of different sizes and using the SATD function as the rate-distortion function, and the generated SI frame retains more detailed information. In addition, we propose an motion estimation (MV) prediction algorithm that further utilizes the motion correlation between adjacent coding units within the video frame on the basis of HEVC-ME. Before the ME, the search starting point is compensated to obtain a more accurate search range to enhance the quality of the SI frame. Experimental results demonstrate that the overall performance of the proposed scheme surpasses that of traditional methods.