Coding with Temporal Layers or Multiple Descriptions for Lossy Video Transmission

Abstract

In this paper, we compare temporal layered coding (TLC), as well as single-state coding (SSC), to multi-state video coding (MSVC) in the context of lossy video communications. MSVC is a Multiple Description Coding (MDC) Scheme where the video is coded into multiple independently decodable streams each with its own prediction process and state. The performance of these three coding schemes are analyzed at different loss rates and coding options, under the assumption that each packet contains the complete coded data for a frame, and the total bit rate is kept constant. To substitute the lost frames, MSVC employs state recovery based on motion compensated frame interpolation, whereas SSC and TLC repeat the last received frame. Results show that MSVC outperforms SSC and TLC for high motion sequences, and also for low motion sequences at high loss probabilities, due to increased state recovery ability of the system. Additionally, if one of the parallel channels of MSVC is in bad condition, unbalanced MSVC that allocates less bit rate to this channel, becomes favorable. Finally, increased error resilience with intra-GOB or frame update improves the system performance for high motion sequences at high loss rates, whereas for low motion sequences, intra updates are disadvantageous due to the penalty on the source coding quality.