Fast Algorithm For Stereoscopic Video Coding Standard 3D-HEVC

3D-HEVC is the new generation of stereo video coding standard.As the extension of HEVC(High Efficiency Video Coding),3D-HEVC still uses the traditional hybrid coding framework and coding technologies the same as HEVC.In addition,adds some new technologies that are beneficial to stereo video coding.Compared with MVC(Multi-view Video Coding)which is the pervious stereo video coding standard based H.264,3D-HEVC can obtain higher coding property.However,the increase of its coding property brings huge computational complexity.Therefore,how to reduce the computational quantity has important value and significance.For that,this thesis develops a series of optimization algorithms for depth Intra prediction and Inter prediction which have the superior time-consuming parts in 3D-HEVC.For accelerate Intra prediction,this paper proposes two algorithms.Firstly,this thesis proposes a fast mode decision algorithm named SOG-FIMD(Sum-of-Gradient Based Fast Intra Depth Mode Decision)after analyzing the correlativity between the smooth degree of coding block and the best prediction mode.When the coding block is smooth enough,some prediction mode can be skipped,because these modes are hardly ever used when the coding block is smoothness.In order to determine whether the coding block is smooth,we use the sum of Roberts gradient of current coding block as criteria.If the sum of gradient of a coding block is smaller than a threshold T,then consider current block as a smooth block.Experimental results have shown that compared with the original algorithm,the proposed algorithm can reduce encoding time by 16.17% on average,while bringing only 0.01% BD-rate loss on synthesis views.Secondly,a fast partition algorithm named SOG-FCUP(Sum-of-Gradient Based Fast CU Partition)is presented in this paper by analyzing the rate distortion optimization function used in process of coding unit partitions.Reduce computational complexity through directly terminating the further segmentation for the completely flat coding block of the depth map.Experimental results have shown that the proposed algorithm shortens 16.7% encoding time,and reduces 0.01% BD-rate on synthesis views.For accelerate Inter prediction,this paper proposes two algorithms.Firstly,this paper developes a Skip mode early determine algorithm named DSMD(Dependent View Skip Mode Early Determination)after analysing the characteristic of 3D video.Determined the prediction mode of coding block in dependent view,according to the best prediction mode of co-located coding block and adjacent block in independent view.Experimental results have shown that the proposed algorithm is able to achieve a reduction of encode time by 6.50% on average,while incurring only 0.08% BD-rate loss on synthesis views.Secondly,by analyzing the relation between video texture and predicton mode,this paper raises an algorithm about selectively skip checking symmetric motion partition(SMP)mode and asymmetric motion partition(AMP)mode according to the texture features of coding block,named SASK(SMP& Mode Selectively Skip).Experimental results have shown that the proposed algorithm can reduce encoding time by 12.71% on average,while bringing only 0.10% BD-rate loss on synthesis views.In summary,the prpposed algotirhms can effectively reduce encoding time while maintain the same coding efficiency.