Error Concealment For Three Dimensional Video Coding And Transmission

Along with fast development and compact combination of both computer technology andcommunication technology,the conventional two dimensional video coding, transmissiontechnologys are gradually becoming mature. As the flexible interactivity and highly stereo feeling,the three dimensional videos more and more get the favour of people, and constantly attract moreand more industry, academia, as a result the three dimensional videos are developing to be a mainresearch directions of the multimedia. Because of data quantity of the three dimensional videos areseveral times as much data quantity of ordinary two-dimensional videos, the exiting compressiontechnologys are used to reduce the correlations of time, space and viewpoint. At the same time thefallible current IP network and the limited bandwidths make the video streams fragile. Oncemistakes appear, the lost data will not only affect the current frame and the subsequent frameswhich reffered to the current frame in the same viewpoint, but also the frames which reffered to thecurrent frame in the adjacent viewpoints, as a result the videos will apper mosaic, black blocks andthe motionless frames. In order to resist the unpleasant problems, error concealment for threedimensional video has a important significances. This dissertation focus on the algorithm of errorconcealment for three dimensional video, and the dissertation propsed the correlation mapping andutilize the correlation mapping to adaptively adopte the error concealment methord to recovery thedamaged frames. In this dissertation the main research results obtained are as follows:(1) Research and analysis of the current three dimensional video error algorithms. The currentthree dimensional video error algorithms have beening attracting attention of the industry andacademias, the dissertation researches the current three dimensional video error algorithms.(2) Exploited an error concealment platform based on the three dimensional videoencoding/decoding test software. There are no error concealment codes in the internationalstandard for multi-view video decoding test software, an error concealment platform based onmulti-view video encoding/decoding test software was exploited, it has a great significance.(3) Proposed the correlation mapping for three dimensional videos, and set the important stagefor adaptively choosing the error concealment method. The three dimensional videos not onlyinclude temptal and spatial correlations of the conventional two dimensional video, but also includeviewpoint correlation of adjacent viewpoint. The rich correlation of the three dimensional videosunavoidably have the different effects, the general video recovery methods are in turn to use allsorts of video correlation to get the candidate frames which are used to replace the damaged frames, then choose the best concealed frames from the damaged candidate frames by an evaluationcriterion or weighted summation as the finally recovered frames. The above methods are aimless inthe choice of the correlation, and can not avoid to lead to increases computational complexity.Thisdissertation proposes a three dimensional videos correlation mapping, by the mapping the besteffect of the damaged frame‘s correlation can be prected before caluateing the candidate frames.Experiments prove the correlation mapping can not only simplify the computational complexity,but also can make better video effect.(4) Proposed an error concealment algorithm for multi-view video based on correlationmapping. Resently the exiting error concealment methords are limite to situation of losting partblocks of a frame. Because there are no correct received blolcks adjacent the lost blocks, themethords of losting part blocks of a frame cannot simply used to the losting whole frame. Thealgorithm proposes an error concealment algorithm for multi-view video based on correlationmapping, which utilize the correlation mapping to adaptively choose the corresponding errorconcealment methord from the candidate error concealment methords sets which include temptal,spatial and viewpoint error concealments to recove the damaged frames. Experements prove thatthe proposed can get better video subjective and objective quality. In the siuttion of different QP,the proposed algorithm increases average13.40dB than the time frame copy algorithm, average12.57dB than the temptal motion exportion, and average9.57dB than the linear interpolationbetween the viewpoints in the same QP.