End-to-end Distortion Modeling and Channel Adaptive Optimization of Mixed Layer Multiple Description Coding Scheme

Multiple Description Coding (MDC) is an error resilient coding technique in which multiple streams from a video source are generated, each is individually and mutually decodable. This error resiliency is obtained at the cost of redundancy, and the amount of redundancy depends on the channel loss rate as well as the frame position in the sequence. Due to the nature of video codecs, an erroneous frame will infect the successive frames and causes an unacceptable quality. MDC redundancy allocation more than required leads to compression inefficiency. Therefore, a channel adaptive optimization for frame-wise redundancy allocation is inevitable. In this paper, the MDC scheme known as Mixed Layer Multiple Description Coding (MLMDC) is under consideration for end-to-end distortion modeling and channel adaptive optimization. The model works based on the side and central decoder outputs mismatch. The performance of the model as well as the optimizer are verified by experimental results measured from JM19.0, H.264/AVC reference software. The experiments also show that the optimal MLMDC outperforms the conventional methods for high enough loss rates.