Side Information Generation for Multi-view Distributed Video Coding Using Hybrid Search and Multi-selection

The quality of side information (SI) has a direct impact on the rate-distortion (R-D) performance of multi-view distributed video coding (MDVC). In this paper, we propose an inter-view SI generation scheme for MDVC, which combines hybrid search and multi-selection. In the presented scheme, both feature matching and epipolar line search are employed to improve the accuracy of frame matching. Moreover, multi-selection has been used to obtain the optimal view disparity vectors (DVs). Experimental results show that under the same coding bitrates, the reconstructed video quality of the proposed scheme is superior to that of the method employing disparity compensated view prediction (DCVP) technique significantly. Additionally, without knowing the camera intrinsic parameters, the presented scheme can achieve the same decoded video quality compared with the approach using disparity based view synthesis (DBVS) technique.