分布式视频编码中基于多概率混合分布的相关噪声建模方法

相关噪声建模(Correlation Noise Modeling,CNM)的准确度是影响分布式视频编码(Distributed Video Coding,DVC)系统性能的关键因素之一.针对经过8×8离散余弦变换(Discrete Cosine Transform,DCT)后的相关噪声子带,本文提出了全新的基于多概率混合分布的CNM方法.该方法可根据子带的信息熵与子带在柯西、拉普拉斯和高斯三种概率分布下信息熵的相似度,自适应选择合适的概率分布对子带进行建模,提高了CNM的准确度.实验结果表明,采用本文提出的离线和在线CNM方法与现有典型的CNM方法相比,DVC系统的率失真(Rate-Distortion,R-D)性能均获得显著的提高.

Correlation Noise Modeling Algorithm Based on Multiple Probability Distributions for Distributed Video Coding

Accurate correlation noise modeling (CNM) is one of the key factors affecting the performance of distributed video coding (DVC).This paper presents a novel CNM algorithm based on multiple probability distributions for DVC,which mainly focuses on the subbands of correlation noise after 8×8 discrete cosine transform (DCT).The proposed CNM method can select the best suitable probability distribution to model the subbands of correlation noise adaptively by comparing the entropy of the subbands with the ones of the candidate probability distributions including Cauchy,Laplace and Gaussian.Experimental results show that compared with the existing typical CNM approaches,the proposed CNM method can improve the rate-distortion (R-D) performance of the DVC system significantly both in offline and online manner.