基于视差调整的3D视频重绘方法研究

项目名称： 基于视差调整的3D视频重绘方法研究

项目编号： No.61002043

项目类型： 青年科学基金项目

立项/批准年度： 2011

项目学科： 金属学与金属工艺

项目作者： 朱云芳

作者单位： 浙江工商大学

项目金额： 7万元

中文摘要： 本项目提出了基于视频调整的3D视频重绘方法研究。根据人眼的深度感知机理，感知深度的变化最终要通过（屏幕）视差的变化来体现。为了适应不同屏幕尺寸对人眼感知深度的影响以及减轻由于辐辏和调节之间的矛盾导致的视疲劳现象，在本项目中我们提出了对立体视频的视差信息进行主动调节进而对观众的视觉深度体验进行调整的立体视频重绘方法：输入待回放的立体视频，得到原左右视图，通过立体匹配求取视差信息，根据人眼对深度的视觉认知特性对视差进行调整，再根据调整后的视差对左右视图进行重绘，经过后处理后输出。在此基础上，本项目重点研究了以下基本问题：（1）立体显示系统和景深调节方案；（2）时空结合的立体匹配方法；（3）视频中运动物体的分割；（4） 视频中感兴趣区域的提取；（5） 多视角几何和摄像机标定。通过本项目研究，共发表学术论文四篇，其中SCI收录1篇（IF 3.72），EI收录2篇，国内核心刊物1篇，已达到预定的研究目标。

中文关键词： 3D视频；深度感知；重绘；视差；立体匹配

英文摘要： This project studies the mechanism and realization of 3D video re-rendering methods based on adjustment of the disparities. Learning from the congnition of human vision system, when people viewing the 3D video on the stereo displaying screen, the disparity on the screen are the key factor affecting the depth information perceived by the people. In order to adapt to different sizes of stereo screen when displaying 3D videos and relieve the visual fatigue caused by vergence-accommodation confilicts, we propose to re-generate a new 3D video according to the depth information computed from disparities, which will affect the 3D experience of the audiences. The main processes are as follows: input a stereo video, get the left frame and right frame, extract the disparity by stereo matching, adjust the disparity according to the cognition of HVS, repaint the new left and right frame based on the modified disparity values, after post-processing, output the new left and right frame. In the framework of the proposal, several key problem are studed: 1) the stereo displaying system and method of depth adjustment; 2) spatial-temporal consistent stereo matching; 3) segmentation of moving objects in video; 4) region of interest extraction in video; 5) multi-view geometry in computer vision and calibration. We have published 4 papers, in which 1 paper is SCI indexed and 2 papers are EI indexed. The research objectives have been reached.

英文关键词： 3D video;depth perception; video rendering;disparity; stereo matching