Practical distributed video coding in packet lossy channels

Abstract

Improving error resilience of video communications over packet lossy channels is an important and tough task. We present a framework to optimize the quality of video communications based on distributed video coding (DVC) in practical packet lossy network scenarios. The peculiar characteristics of DVC indeed require a number of adaptations to take full advantage of its intrinsic robustness when dealing with data losses of typical real packet networks. This work proposes a new packetization scheme, an investigation of the best error-correcting codes to use in a noisy environment, a practical rate-allocation mechanism, which minimizes decoder feedback, and an improved side-information generation and reconstruction function. Performance comparisons are presented with respect to a conventional packet video communication using H.264/advanced video coding (AVC). Although currently the H.264/AVC rate-distortion performance in case of no loss is better than state-ofthe-art DVC schemes, under practical packet lossy conditions, the proposed techniques provide better performance with respect to an H.264/ AVC-based system, especially at high packet loss rates. Thus the error resilience of the proposed DVC scheme is superior to the one provided by H.264/AVC, especially in the case of transmission over packet lossy networks.

Scalable Distributed Video Coding for Wireless Video Sensor Networks

“…The WZFs obtained different bit rates due to different sub-band combinations on different layers after wavelet transformation to construct the BL, and three ELs (EL 1 , EL 2 , EL 3 ). All the ELs in the SDVC are encoded by the Low Density Parity Check (LDPC) codec, which is proved to be better channel codec in packet lossy network for DVC than LDPC accumulated (LDPCA) code [19].…”

Context-Adaptive Modeling for Wavelet-Domain Distributed Video Coding

“…Because motion-compensated prediction accounts for most of the computational load, DVC can shift most of the computational load from the encoder to the decoder. Meanwhile, DVC has been proven to have strong error resilience, 3 thanks to it having no prediction loop at the encoder. These unique features have made DVC a good solution for scenarios where the encoder has limited processing capability and power supply but where the decoder is relatively powerful-for example, in multimedia wireless sensor networks.…”