双目立体数据的高效获取和智能处理技术研究

In recent years, the development of academic research and applications of visual media demonstrates a significant trend toward 3D. 3D media with its powerful expression and vivid visual experience has quickly attractted people's attention. Compared to traditional 2D images and video, there are still relatively fewer binocular stereoscopic media resources, and editing and processing technologies are not mature. These are two main difficulties for binocular stereoscopic technology. How to efficiently acquire, create, process and present the high quality binocular stereoscopic data is an important problem which has extensive application prospect and great theory value. .This project is focusing on binocular stereoscopic data, mainly study the following aspects: the analysis of visual distortion in real-time stereoscopic acquisition and binocular viewpoint optimization, intelligent 3D conversion of 2D media based on 3D structure estimation, the vector format of binocular stereoscopic data and interactive creation technology, free stereoscopic touring for stereoscopic media, seamless blending of stereoscopic data based on gradient domain analysis, intelligent stereoscopic data inpainting based on fast region matching and the complementary of binocular stereoscopic data, efficient and intelligent processing system for binocular stereoscopic data and the design of stereoscopic acquiring and presenting system prototype.

近年来，可视媒体的研究和应用开始向3D发展，3D媒体凭借其强烈的表现力和逼真的视觉体验迅速成为人们关注的热点。和传统的二维图像和视频相比，双目立体媒体资源相对稀缺，相关处理技术不够成熟，是当前双目立体技术面临的主要问题。如何能够高效的获取、创建、处理并呈现高质量的双目立体数据具有重要的研究价值和应用前景。本项目针对双目立体数据，着重开展如下方面研究：双目实时采集中的视觉扭曲度分析和双目视点优化，基于三维结构估计的二维数据智能双目立体化，矢量格式的双目立体数据交互式创建技术，高自由度的双目数据视点漫游，基于梯度域的多源双目数据高效无缝融合，基于区域快速匹配和双视点互补分析的双目立体数据智能补全，面向双目立体数据的高效智能处理系统和双目采集与成像系统原型设计。

近年来，可视媒体的研究和应用开始向3D发展，3D媒体凭借其强烈的表现力和逼真的视觉体验迅速成为人们关注的热点。和传统的二维图像和视频相比，双目立体媒体资源相对稀缺，相关处理技术不够成熟，是当前双目立体技术面临的主要问题。本项目研究双目立体数据的高效获取和智能处理技术，针对双目立体数据，研究了双目立体数据的认知计算模型，以及视点调整、舒适度度量方法和基于舒适度的双目视频编辑、显示优化等算法，研制了高效的双目视频采集、成像与传输系统，实现远程呈现和交互。项目发表论文10篇，其中6篇发表在ACM Transactions on Graphics、IEEE Transactions on Visualization and Computer Graphics、IEEE Transactions on Image Processing等著名期刊上，申请发明专利4项，2项已经授权。