Optimized Bit Extraction of H.264-SVC MGS stream over error-prone channels

Abstract

Classical Bit Extraction algorithms can guarantee truncated scalable coded video stream with good reconstructed quality under error-free bandwidth limited channels. However, they are deficient in two folds: firstly, packet loss is not considered in the truncated rate-distortion model; secondly the reference of original video sequence is unavoidable. To overcome the above shortcomings, this paper firstly introduces a low complexity error propagation distortion model, in which error drift distortion due to rate truncation and packet loss can both be analyzed in the absence of the original sequence. Based on this model, an optimized bit extraction algorithm is proposed for the transmission of H.264/SVC MGS stream over bandwidth limited error-prone channels. Extensive simulation results show that, the error drift distortion of both base layer and enhancement layer can be accurately predicted based on the introduced error propagation distortion model. In addition, compared with the benchmark bit extraction algorithms provided by JSVM, our proposed algorithm can not only provide a higher reconstructed quality of truncated MGS video stream under various rate constraints and channel conditions, but as well achieve the promising performance with a much lower time complexity.