Temporal error concealment algorithm for H.264/AVC using omnidirectional motion similarity

Abstract

H.264/AVC is the newest one among several video compression standards. The main goals of H.264/AVC are to achieve efficient compression performance and a network friendly video coding. However, if an error occurs when transmitting compressed video, error concealment is needed to prevent error propagation and to improve the video quality. In this paper, we propose the temporal error concealment algorithm which provides high performance for H.264/AVC. The proposed algorithm uses the property that the motion vectors (MVs) between the error macroblock (MB) and the neighboring MB have high similarity to select a group of candidate MVs, when an error occurs in the inter-coded frame. Next, weighted overlapped boundary matching algorithm using the credibility of information selects the best candidate MV among a group of candidate MVs. The experimental results show that the proposed algorithm improves PSNR up to 3.02 dB compared with the boundary matching algorithm (BMA).