The Research On Side Information Generation Technology For Multi-view Distributed Video Coding

In recent years, Distributed Video Coding (DVC) has attracted more and more attention due toits simple encoder and complex decoder. Multi-view Distributed Video Coding (MDVC) is theextending application of DVC in Multi-view Video Coding (MVC). As in mono-view DVC, sideinformation (SI), which is closely related with the encoding performance, is also a core technologyin MDVC.This paper launched a series of studies on SI generation technology in MDVC, includingintra-view SI, inter-view SI and SI fusion. The main research works are as follows:(1) In-depth study of the classic SI generation method, named Motion Compensated TemporalInterpolation (MCTI) is conducted. In the segment of unidirectional motion estimation, the adaptiveobtaining of the search area is introduced as the computing complexity of the traditional full searchmethod is high and the authenticity of motion vector is low. Before the final bidirectional motioncompensation, a motion vector smoothing method was introduced to further enhance theauthenticity of the motion vectors. Compared with the MCTI, the experimental results show that theaverage PSNR of intra-view SI generated by the proposed method can be improved significantlyimproved.(2)The research of inter-view SI generation method in MDVC is conducted. This paper usesHomography-Compensated Inter-view Interpolation (HCII) to generate inter-view SI. In HCII, theaccuracy and stability of homography are poor when using RANSAC algorithm. Then ahomography calculation method based on Least Squares Method is proposed. Experimental resultsshow that the proposed method can effectively guarantee the accuracy and stability ofhomography’s obtaining, thus improving the quality of the inter-view SI.(3)The research of common SI fusion methods in MDVC is conducted. Based on the analysisof intra-view SI’s characteristics, this paper presents a criterion to judge whether the movement ofthe macroblock is consistent with the whole frame. Then the result is used to determine the maskwhich is used for SI fusion. Experimental results show the proposed SI fusion method is effectivefor videos with severe movement.