Research On Rate Allocation And Side Information Refinement Algorithms In Feedback-free Distributed Video Coding

Distributed Video Coding (DVC) is a new paradigm for video compression. Comparedwith traditional video coding, it introduces the channel coding techniques such as Turbo codein the channel coding and shifts computational complexity from encoder to decoder toachieve high compression efficiency performance. Thus, DVC is well-suited for the wirelessvideo terminal applications where computing power, power consumption and storage capacityare all limited.Feedback-free DVC compression system is suitable for applications of no feedbackchannels. Because there is no feedback channel between encoder and decoder, it decodes onlyonce, reduces the decoding delay and computational complexity. Rate estimation andallocation at encoder are the key techniques to the feedback-free DVC systems, so they arethe research emphasis in this paper.This paper firstly implements the feedback-free DVC system in DCT domain using the Cprogramming language based on TDWZ system framework. Then we focus on rate allocationalgorithms at encoder and side information refinement algorithms at decoder. The main worksand achievements are as follows:1. The accuracy of rate allocation at encoder is a key factor in DVC system performance,the more accurate the rate allocation is, the better the performance of DVC rate distoration is,and vice versa. On the research of encoder rate allocation algorithms, this paper proposed anrate allocation algorithm based on bitplane error probabilities. Simulation results demonstratethat the proposed algorithm can be reasonably completed parity bit streams allocation at theencoder under the condition of effectively reducing the computational complexity, and it canimprove the system rate distortion (RD) performances and Wyner-Ziv frames reconstructionqualities.2. Because there is no feedback channel between encoder and decoder for feedback-freeDVC system, and the system decodes only once at decoder, so maybe some unsuccessfullydecoded Wyner-Ziv frame bitplanes exist, which reduces the system RD performance. Inorder to obtain higher quality decoding Wyner-Ziv frames, this paper proposed a sideinformation refinement algorithm at decoder to improve the quality of side information.Experimental results show that the proposed refinement algorithm can effectively improve theside information quality, therefore, it enhances the quality of Wyner-Ziv frames and system RD performance. In addition, the algorithm has a positive meaning to our further research onrate allocation algorithms at encoder.