Research On3D Video Compression Method Based Depth Information And Region Of Interest

Three-dimensional(3D) video provides unparalleled visual experience, at thesame time, it puts additional stress on the preservation and the transmission of videobecause of the huge amount of information. Under the premise of guarantee thequality of video as far as possible, how to occupy the smaller transmissionbandwidth has become a challenge that must be solved instantly in3D videocompression field. Therefore, there is great need to find a compression and encodingmethod corresponding to3D video features to improve the compression efficiency,at the mean time, we should fully consider the compatibility with the existingcompression standard.The region of interest(ROI) coding ensure the subjective quality of video andeimprove the compression efficiency at the same time by controlling the bit-rateallocation between video background region and ROI. As to2D video, ROI andbackground region automaticly selection is difficult to achieve, which has limitedthe application of the ROI compression. Whereas, there is a strong link betweenscene depth information contained in3D video and the degree of interest in humanvisual model (HVS), which provides conditions for3D video ROI automaticallyselection.This thesis mainly studies with the ROI compression method of the2D videoplus depth map based on the depth information, which is compatible with H.264standard. First of all, the stereo vision model of human beings and the relationbetween the depth and the degree of interest in HVS are analysed in this study. Thus,the ROI automatically selection method is proposed, which makes use of the depthmap segmentation based on mean shift and edge extraction based on sobel operator.Afterwards, the paper discusses the compression process of H.264standard,particularly the constitution and control mode of the bit stream. Accordingly, theROI pretreatment means by spatial filtering and spatial temporal filtering in thebackground is determined. Finally, the effects to the bit rate after compression ofthese two pretreatment methods are assessed through both experiment and theory.The compression experiment was performed with JM10.2encoder in H.264standard.The results have proved the feasibility of3D video ROI compression method in thispaper. The bit rate of compressed data stream has reduced by about15%and thePSNR in ROI maintains almost the same.The technology of3D video ROI compression based on the depth informationowns great value in research and application. It utilizes the relevance of the depthand the degree of interest, at the same time, it is capable of favorable compatibility with2D video compression standard. In addition, it is conveniency to extend thistechnology to the advanced3D video compression format.