分散式視訊編碼在封包遺失通道的錯誤復原機制

Abstract

Low-power and low-cost distributed wireless video sensors play important roles for applications in machine-to-machine (M2M) and wireless sensor networks. Conventional video coding standards consume large power consumption at the encoder side.This kind of video coding systems do not meet the requirements of M2M or wireless sensor networks, where the hardware resources of the sensor nodes are limited. On the other hand, Distributed video coding (DVC) based on Slepian-Wolf and Wyner-Ziv theories provides a promising solution for implementing low-power and low-cost distributed wireless video sensors since most of the computation load is moved from the encoder to the decoder. For lossy channels in the real-world environment, the error resilience schemes against packet loss for conventional video coding have been extensively studied. However, for DVC, the effect of packet loss and the error resilience scheme has not been fully investigated. In this thesis, the effect of packet loss is analyzed, and the error resilience schemes for DVC are then developed. Since the two types of frames in DVC, Key frames and Wyner-Ziv frames, have different characteristics, we discuss the effect of packet loss over key frame channels and Wyner-Ziv channels, respectively. For lossy key frame channels, three resilience schemes can achieve a PSNR gain of 2 to 8 dB compared to the baseline. In addition, the proposed hybrid mode can select mode successfully to achieve better rate distortion performance. For lossy Wyner-Ziv frame channels, two error protection schemes, forward error protection and additional source packets are applied to Wyner-Ziv packets. The experiments show that these two schemes are not effective, since the Wyner-Ziv packets inherently have good error resilience ability under the DVC architecture with iterative requests. For the condition that key channels and Wyner-Ziv channels are both lossy, based on the experimental results of rate allocation, we also propose an error resilience scheme for the whole DVC system. The proposed error resilience scheme effectively reduces the PSNR gap between ideal case and baseline and improves around 0.5 dB in PSNR for sequences Hall and News compared with the other two resilience schemes.