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

Depth map was introduced in 3D-HEVC to achieve better view synthesis.According to the characteristics of depth map,depth modeling modes(DMMs)were added in the mode selection process of the depth map in 3D-HEVC.In addition to the 35 traditional modes,there are DMMs of intra mode in 3D-HEVC,such a large number of modes lead to a substantial increase in coding time.To solve above problem,and the experimental results show energy concentration property of DCT can obviously distinguish whether the coding block has the edge in depth map encoding process of 3D-HEVC.The paper proposes DCT based intra mode selection algorithm for depth-map coding in chapter 3.The method simplifies the DCT transform,ignores the lower right corner coefficient and AC coefficient of DCT,only 6 coefficients from the top left corner which are the most discriminative are calculated.The experimental results show that the energy concentration degree of each coefficient is different,thus the 6 coefficients have different weights.According to the weights,the coefficient value and the threshold value are compared in order,if the given condition is satisfied,DMMs are added to the candidate list;Otherwise,DMMs are not added,and do not calculate RD-cost.Experimental results show BD-rate is reduced by 1.78%,BD-PSNR is increased by 0.21 dB and average encoding time is reduced by 34.52% compared to reference algorithm.The main idea of DCT based algorithm is to add DMMs to the candidate list or not,further study found that each mode of DMMs can be applied to the depth image with different edge complexity,mode selection can be more refined,so edge complexity based intra mode selection algorithm for depth-map coding is proposed in chapter 4.This algorithm first detects image edge points,then calculates the edge direction of each edge point,and the direction normalization is used as the edge complexity.Comparing given threshold and normalized results,if result shows the picture has large edge complexity,adding mode in DMMs suitable for complex edges to candidate list only;otherwise only adding mode suitable for simple edges of DMMs to candidate list.All DMMs do not add to list if picture has no edge.Experimental results show BD-rate is reduced by 0.09%,BD-PSNR is increased by 0.09 dB and average encoding time is reduced by 29.24% compared to reference algorithm.In order to apply these two fast algorithms in practice,we realized the system platform of video capture and encoding/ decoding with array camera and introduce in chapter 5.The system platform consists of two parts: Hardware(array camera and PC)and software(camera,configuration,recording and encoding,decoding unit).It realizes multi view video capture,recording,encoding and decoding.The experimental results show that the optimization performance of the algorithm achieves the same level in the application of the system platform.