Adaptive error-resilient techniques for H.265/HEVC video transmission

Abstract

High efficiency video coding (HEVC) is the latest video coding standard that enables the transmission of high resolution videos over wireless networks. In this dissertation, HEVC based error resilient video coding is proposed to target real time video transmission applications. A feedback-based reference picture selection (RPS) method called hierarchical-P RPS is proposed such that it can adapt with the HEVC coding framework. Due to hierarchical coding structure of HEVC, the quality of each frame in the sequence varies depending on its location in the hierarchical structure. High quality pictures are used as reference for several subsequent pictures while low quality pictures are used for only one consecutive picture. When error is detected, hierarchical-P RPS algorithm will select suitable reference pictures subject to the trade-off of low quality picture with short temporal distance and high quality picture with long temporal distance. To enhance the performance of the proposed RPS, we combine feedback-based RPS and region-of-interest (ROI) based intra refresh method by using some novel features of HEVC. Each block in the ROI region is encoded with intra mode according to error status such that it can enhance the quality of picture in ROI area and can avoid the motion search process in this area. To enable block level intra refresh, rate control is also modified to fit in the framework. Experimental results demonstrate that the proposed framework can achieve better subjective and objective quality, compared to the hierarchical-P RPS algorithm. The average PSNR improvement of proposed algorithm over modified RPS algorithm is 1.04 dB and that over HEVC reference software with regular intra frame refresh is about 5.23 dB.