3D-HEVC视频编码技术研究

发布时间：2018-04-17 03:17

本文选题：HEVC + 3D-HEVC　；参考：《西安电子科技大学》2014年硕士论文

【摘要】：HEVC是ITU-T和ISO联合制定的下一代视频编码标准，相比于H.264/AVC标准，在重建质量相同的情况下，HEVC可以节省约一半的码率。3D-HEVC是HEVC的扩展。作为一种基于HEVC的新3D标准，3D-HEVC以多视点视频加深度为压缩数据格式，用于支持立体和交互的视觉体验。本文主要研究3D-HEVC中的编码技术，包括视频图像的视差矢量预测技术和深度图预处理编码技术。视差矢量预测通常采用的方法是利用时域和空域相邻块的视差信息估计当前编码块的视差信息。但当物体运动时，该方法估计的视差信息精度不高，影响了视差矢量使用的概率及视点间预测的精度。为此本文提出利用当前编码块的运动信息确定对应参考块在参考帧中的位置，提取对应参考块的视差矢量信息，用于预测当前编码块的视差信息。通过该方法，提高了获得的视差信息精度，增大了视差矢量预测编码模式被采用的概率，降低了编码所需码率。深度图编码是3D编码的重要组成部分。目前深度图主要通过估计算法获得。由于估计算法精度的限制，，大部分深度图的相邻帧背景区域存在闪烁现象，不利于进行编码压缩。本文利用视频图像的帧差图估计出背景区域，并对该区域所对应的深度图区域进行平滑滤波，以增强深度图背景区域时间相关性。另外，本文对视频图像帧差图进行腐蚀和膨胀操作以便去除噪点、保护运动前景及边缘信息。实验结果表明，所提算法在不影响合成视点视频质量的前提下，增加了帧间skip和merge模式使用概率，降低了编码码率。
[Abstract]:HEVC is a next-generation video coding standard developed jointly by ITU-T and ISO. Compared with the H.264/AVC standard, HEVC can save about half of the bit rate. 3D-HEVC is an extension of HEVC under the same reconstruction quality.As a new 3D standard based on HEVC, 3D-HEVC uses multi-view video and depth as compressed data format to support stereoscopic and interactive visual experience.This paper mainly studies the coding techniques in 3D-HEVC, including parallax vector prediction of video images and depth map preprocessing coding.Parallax vector prediction usually uses parallax information of adjacent blocks in time domain and spatial domain to estimate parallax information of current coded blocks.However, when the object is moving, the accuracy of the parallax information estimated by this method is not high, which affects the probability of using parallax vector and the accuracy of prediction between viewpoints.In this paper, the motion information of the current coded block is used to determine the position of the corresponding reference block in the reference frame, and the parallax vector information of the corresponding reference block is extracted to predict the parallax information of the current coded block.Through this method, the accuracy of parallax information is improved, the probability of parallax vector prediction coding mode is increased, and the coding rate is reduced.Depth map coding is an important part of 3D coding.At present, the depth map is mainly obtained by the estimation algorithm.Because of the accuracy limitation of the estimation algorithm, there is a flicker in the background region of the adjacent frames in most depth maps, which is not conducive to coding compression.In this paper, the background region is estimated by the frame difference graph of the video image, and the depth map region corresponding to this region is filtered smoothly to enhance the temporal correlation of the background region of the depth map.In addition, the frame difference image of video image is corrupted and inflated to remove noise and protect motion foreground and edge information.Experimental results show that the proposed algorithm increases the usage probability of skip and merge modes between frames and reduces the coding rate without affecting the quality of synthetic view video.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN919.81

【参考文献】