Font Size: a A A

Research On Multiview Coding Technology Of3D Video

Posted on:2013-03-17Degree:MasterType:Thesis
Country:ChinaCandidate:Z H LiFull Text:PDF
GTID:2248330395962847Subject:Computer application technology
Abstract/Summary:PDF Full Text Request
Because of the real, natural and broad application prospects,3D video is widely used in3D movies,3D TV,3D video conference, as well as military, medical and so on. The amount of data of3D video is so large that it is difficult for the existing system to meet the demand for storage, processing and transmission, so how to effectively encode video information is the focus of the study. This paper studies the multiview coding framework, video coding prediction structure and depth map-based synthesis method, effectively reduce information redundancy and improve coding efficiency.This paper first studies multiview coding technology of3D video, then explores the multiview coding frame which based on H.264/AVC, improves the multiview prediction structure and the virtual view synthesis method, and they are verified effectively. The research work are as follows:one, three-dimensional video imaging principle, image coding theory, the key technologies of H.264/AVC video coding as well as the multiview coding evaluation criteria. Second, multi-view’s forecasting techniques have important implications on video compression, after analyzing the Simulcast prediction, sequential perspective prediction and viewpoint/time hybrid prediction based on the hierarchical B frame, this thesis proposes an improved prediction structure, which based on the MVC reference predict structure and made full use of the GoGop predictable ideological. The experimental analysis shows that this prediction can improve random access performance of video and reduce the encoding complexity, and at the same time no significant reducing on the video coding efficiency. Third, on the virtual view synthesis, a method is proposed,which based on the image fusion of DIBR. And this method adopts weight of the mean filter to smoothing preprocessing, which use a3×5weights templates. After three-dimensional image mapping, this method uses the image fusion methods for hollow’s further processing. The test proved that this synthesis method is better than the old one.
Keywords/Search Tags:3D video, multiview coding, multiview prediction structure, DIBR
PDF Full Text Request
Related items