Study On Theory And Algorithms For Stereo Image And Video Coding

In recent years, stereo image and video coding have become a novel research area in the field of video signal processing. The main purpose of stereo image and video coding is to effectively reduce data making use of some characteristics of stereo image and video, which will result in easily storing and transmitting. The key techniques include disparity'estimation, occlusion detection, disparity compensated difference coding and joint estimation of disparity and motion. In this dissertation, research emphases are placed on some of the aspects motioned before and several efficient algorithms are proposed. Furthermore, the future trends of this research direction are proposed.The dissertation proposes a novel adaptive block-size stereo image coding algorithm based on traditional block-based coding algorithm, which fully utilizes the information of intensity variety. It can adaptively select appropriate matching window to estimate disparity vector in terms of content of image, which improves both the veracity of disparity estimation and coding efficiency of system in some extent. In addition, the feature point-based disparity estimation method is introduced into stereo image coding by making use of Delaunay Triangular mesh model. It not only improves the veracity of disparity estimation but also well employs some advantages of triangular model on image coding. In order to further improve coding efficiency of system, some special algorithms including disparity vector optimization algorithm, occlusion detection algorithm and new vertex transmitting order, are designed for triangular mesh, which leads to better performance on both subjective and objective aspect. Furthermore, the dissertation analyses the characteristics of stereo error image from both mathematics analysis and data statistic, then points out that DCT is more benefit to stereo image coding than the wavelet transformation is. Based on this conclusion, a DCT-based embedded stereo coding algorithm is proposed. Its entropy coding uses improved bit-plane coding method that not only improves coding efficiency but also provides embedded bit streams.This dissertation classifies conventional stereo video coding algorithms into three types according to different prediction modes, and describes them in detail. Then, a novel stereo video coding algorithm named multi-modes prediction is proposed. It selects optimal mode from separate motion estimation, separate disparity estimation and joint motion and disparity estimation using special R-D cost function, which largelyimproves coding efficiency of the system. At last, the dissertation points out that scalable stereo video coding will become an active research topic in the future after studying existent scalable video coding algorithm, such as FGS and PFGS. Based on the products of scalable video coding, two scalable coding frameworks for stereo video and their corresponding rate allocation schemes are introduced. In this part, the author would like to give some elementary suggestions of this topic, which provides the future development of this research direction with a meaningful reference.