Research On Fast Algorithm Of 3D Video Coding Based On HEVC Standard

With the rapid development of video coding technology and display technology,audiences require more realistic visual experience for video,so Free Viewpoint Video(FVV),which can support stereoscopic display,has already become current research topic.In July 2012,3D extension of High Efficiency Video Coding(3D-HEVC)standard was developed by Joint Collaborative Team on 3D Video Coding(JCT-3V).3D-HEVC standard adopts advanced HEVC standard to encode single view texture video and depth map in interpicture and intrapicture,as well as employs Multiview Video and Depth(MVD)format to synthesize virtual views and achieve inter-view coding.HEVC standard has optimized the inter coding,intra coding and so on,as well as MVD format has put forward the conception of inter-view prediction,so that the coding performance of 3D-HEVC standard is high efficient consequently.Especially for intra prediction of texture video and depth map,the accuracy of them has been increased on account of additional 33 intra direction prediction modes and Depth Modeling Mode(DMM).However,adding these new coding tools leads to high computational complexity of multi-view texture video and depth map in intracoding procedure.Therefore,this paper proposes two approaches to overcome these issues efficiently.The first method of this paper utilizes Sobel operator to convolve with Prediction Unit(PU),which can find the edge direction in PU.Thus,the intra direction prediction mode in PU can be determined based on the edge direction.This algorithm avoids intrinsic Rate-Distortion Optimization(RDO),so it accelerates intra prediction mode decision for texture video significantly.Moreover,the average coding time 39.6% is saved by this algorithm while maintaining nearly the same performance as traditional HEVC encoder.The second method of this paper uses OTSU's method to determine Wedgelet mode in depth map coding.The OTSU's method can work out threshold rapidly for partitioning PU in depth map.Our proposed algorithm takes place of conventional DMM determination,thus the calculation complexity of intra prediction mode decision in depth map is decrease dramatically.Furthermore,the average coding time 35.5% is reduced by the second algorithm without obvious loss of coding performance.The two proposed methods in this paper have fulfilled reduction for computational complexity in intra prediction of video coding,and achieved the expected effect,which facilitates the real-time applications of 3D-HEVC standard.