Research On Fast Algorithm For Depth Video Coding

The emerging HEVC standard improves nearly 50% coding efficiency comparing with H.264/MPEG-4 AVC standard.As a 3D video coding standard,3D-HEVC inherits the coding techniques of HEVC and introduces many coding tools for depth video coding.However,the improvement of coding efficiency results in the increase of computational complexity,which hinders the application of 3D-HEVC standard.It is important to research fast coding algorithms based on 3D-HEVC.This paper introduces the video compression coding and the development of video coding standards.Then it elaborates the coding frameworks and key technologies of HEVC and 3D-HEVC standards.Based on the analysis of the characteristic for depth video,the paper focuses on the fast algorithms for depth video coding to reduce the computational complexity.In order to ensure the performance of depth map coding,a new depth intra coding tool called Depth Modeling Mode(DMM)is introduced in 3D-HEVC.However,the process significantly increases the computational complexity due to the blind traversal of all wedgelet partitions.In this paper,a fast intra prediction algorithm for DMM mode based on angle constraint is proposed.In the first step,the unnecessary DMM mode is skipped by judging whether the mode with minimum HAD cost in Rough Mode Decision(RMD)is DC mode.In the second step,the direction information represented by a permitted angle range is acquired based on the mode achieved in first step.In the third step,a searching subset is obtained based on the direction information and the position information represented by the coordinates of the pixel with biggest depth value change in each boundary of PU.Then the patterns within the searching subset are tested by view synthesis optimization(VSO)to find the minimum distortion partition.The experimental results show that the proposed algorithm significantly reduce the complexity of DMM mode prediction processing.Texture video owns more abundant feature information than depth map.Depth map is mainly characterized by large smooth areas and sharp edges.Due to the different characteristic between texture video and depth map,this paper presents a fast intra prediction method based on content feature analysis.First,based on the similarity between the separation line in the best matching pattern and the edge in an actual depth PU(prediction unit),a simplified search algorithm for explicit wedgelet signalization mode is proposed to predigest the complex search process.Then,an intra-mode decision strategy by analyzing the Hadamard cost and variance is put forward.Next,by obtaining the coding information of the neighboring coded PU,the current PU partitioning mode is early determined to optimize the prediction process.Finally,an early decision on CU termination strategy is designed by exploiting the spatial correlation.Experimental results show that considerable encoding time saving is achieved with negligible performance loss.