Research On Fast Algorithm For Inter Coding Of Quality SHVC

With the advent of the Big Video era,video technology has appeared in our lives in various forms such as video on demand,live video,video surveillance and video conferencing,and is quietly changing the way people live.In order to deal with the network heterogeneity and user terminal differences,JCT-VC has released the HEVC Scalable Video Coding standard(SHVC).As a scalable extension of the High Efficiency Video Coding(HEVC),the SHVC encoder needs to encode multiple HEVC layers with Inter-layer predictions,which causes the significant increase in coding complexity.This thesis will make some researches and related optimization works on fast coding techniques for quality SHVC.Firstly,an enhancement layer(EL)algorithm based on coding unit partitioning and residual distribution is proposed.In order to develop the algorithm,we have conducted extensive experiments to investigate the relationship for the current coding unit(CU)and neighboring CUs around.According to the experimental statistical results,the fast algorithm is designed to skip larger or smaller depth and reduce the traversal range of CU depth.In the smaller CU depth traversal range,a statistical test method is introduced to examine whether residual coefficient obey a Gaussian distribution to terminate the CU partitioning early.In RA configuration,experimental results demonstrate that the proposed algorithm can save an average of 69.68% and 36.87% EL coding time and total coding time.Correspondingly,the loss of average PSNR and the increase of bitrate are 0.02 dB and 0.94%,respectively.Secondly,a fast algorithm based on depth weight prediction and Inter PU partitioning is proposed.The method defines the depth weight of Inter-layer and the current CU,and selects the depth level of the to-be-encoded CU according to the depth weight.On the other hand,the method skips the unnecessary PU partitioning by using the information of already coded CU.On the basis of fast CU depth level decision,the optimal PU partitioning mode is determined.As a result,the proposed methods can reduce the computational complexity significantly.In RA configuration,experimental results demonstrate that the proposed algorithm can save an average of 60.47% and 31.02% EL coding time and total coding time.Correspondingly,the loss of average PSNR and the increase of bitrate are 0.03 dB and 0.40%,respectively.