Motion-adaptive quantization and reconstruction technique for distributed video coding

Distributed video coding is a new paradigm based on two information theoretical results by Slepian-Wolf and Wyner-Ziv. The architectures designed so far have invariably made use of the uniform scalar quantization schemes along with a few attempts to make the schemes more adaptive. Quantization is one of the major contributors to the large performance gap between conventional video coding standards and distributed video coding. In this paper, an attempt is made to improve the performance of the Wyner-Ziv video coding by making the quantization algorithm more adaptive to the motion content of the video sequence without significantly increasing the encoder complexity. The proposed method also exploits the temporal correlation to provide for online correlation noise classification. Hence, the improved reconstruction technique which uses the correlation noise information is more adaptive to the motion content. Simulation results show that the proposed motion-adaptive quantization and reconstruction technique achieves improved rate-distortion performance.