Research On Fast Algorithm Of 3D-HEVC Depth Map Video Coding

With the advancement of the times,3D video is applied to real life.The most advanced 3D-HEVC is an extension of the high-performance video coding standard HEVC,which achieves a more efficient compression ratio.3D-HEVC inherits the quadtree partition structure of HEVC,and introduces DMMs for sharp edge coding,which greatly improves the coding efficiency,but also increases the coding time complexity.In order to reduce the video coding time,this paper studies two aspects:depth map intra prediction mode decesion and the depth map coding unit size decision.In order to reduce the coding time complexity caused by more intra prediction modes,a fast mode selection algorithm based on DCT is proposed.The algorithm uses the lower right corner of the DCT coefficient matrix to determine whether the coding unit of the depth map contains sharp edges.First calculate the DCT coefficient matrix of the image.Then calculate whether the high-frequency coefficients of the lower right corner of the DCT coefficient matrix are all zero.If the algorithm condition is satisfied,the DMMs are not added to the pattern candidate list,otherwise they are added.Finally,the full rate distortion value in the pattern candidate list is calculated and the optimal mode is selected.Compared with the HTM 13.0 platform algonthim,the BD-rate is increased by 0.9%and the average encoding time is reduced by 8.7%.In order to reduce the coding time complexity caused by the excessive number of quadtree partitions in the intra-frame prediction of depth map,a fast segmentation algorithm based on DCT,a fast segmentation algorithm based on mean difference and a depth map based on variance are proposed.Segmentation algorithm.In the DCT-based coding unit fast segmentation algorithm,the coding unit is divided by the characteristics of DCT energy aggregation,but it has been experimentally proved that due to the flat region The degree of dispersion of pixel values is relatively low,so DCT cannot be used to determine the flat region,and the average difference and variance can well discriminate the dispersion of the data set.In the fast split algorithm of the coding unit based on the average difference,the average difference of the coding units is calculated,and if the average difference of the coding units is smaller than the threshold,the coding unit is not divided,otherwise,the algorithm is compared with the HTM.With the 13.0 platform algorithm,BD-rate increased by 1.0%,and the average encoding time was reduced by 29.8%.In the variance-based coding unit fast segmentation algorithm,the variance of the coding unit is calculated.If the variance is less than the threshold,the segmentation is terminated early,otherwise the coding unit is further divided downward,and the algorithm is relative to the HTM 13.0 platform.The algorithm,BD-rate increased by 0.6%,and the average encoding time was reduced by 52.5%.