Efficient error detection, localization, and concealment algorithms based on enlarged macroblock for error-robust video transmission

Abstract

The video transmission over unreliable channels, such as air and Internet, becomes increasingly important for the better quality of services. In this study, efficient error detection, localization, and concealment methods based on source level redundancies are proposed for the robust video transmission. For the purpose, an error detection algorithm based on an embedded constraint is proposed. Most video compression standards adopt entropy coding which is susceptible to bits errors causing serious video quality degradation. To alleviate this problem, at the encoder side, the algorithm imposes a constraint on a motion vector with negligible degradation of coding performance. Then, at the decoder side, the imposed constraint is examined. The proposed error detection algorithm enhances motion error detection capability over the conventional one. Moreover, a modified motion estimation algorithm is proposed for error localization and temporal error concealment in transmitting videos over unreliable channels. In order to achieve good error concealment performance, the proposed algorithm implicitly imposes spatial correlations on motion vectors by extending the block size and overlapping blocks in motion estimation. Thereby, the obtained motion vectors can be used to improve error concealment performance while keeping the encoding efficiency with negligible overhead. In addition, the proposed motion estimation can provide a new error detection measure so that one can maximally utilize uncorrupted data rather than simply discarding all data in a defected packet. Simulation results show that the proposed motion estimation scheme provides significant improvements in error concealment performance over the existing schemes and improves the bit utility over a wide range of error conditions. By using the proposed error detection, localization, and concealment algorithms, the error robustness is drastically improved. Therefore, the proposed algorithms can be widely applicable ...