3DTV中编码效应消除和高质量双目视图生成研究

3DTV is expected as a new broadcasting paradigm that provides a more natural and life-like visual home entertainment experience to users. However, there are bottleneck problems in practical applications such as coding artifacts and the limited bandwidth of current broadcast environments. In this research, we aim to achieve effective coding artifact reduction and high-quality stereoscopic view generation in 3DTV. First, we analyze 3D images (i.e., video-plus-depth), and identify fatal coding artifacts which degrade picture quality and depth perception seriously. Second, we design an effective 3D dictionary based on sparse representation for coding artifact reduction to improve the picture quality and 3D effects remarkably. Finally, we also investigate minimization of dis-occlusion area, image warping & completion, real-time rendering, and glasses-free 3DTV technology for high-quality stereoscopic view generation. As a result, this research is able to overcome the bandwidth limitations of current broadcasting systems and achieve high-quality stereoscopic view generation in 3DTV.

3DTV是广播电视领域的一种新的播放模式，它可提供更加自然、生动的家庭视觉娱乐，其中存在的编码效应和带宽问题已经成为实际应用的瓶颈问题。本项目旨在分析降低图像质量和立体效果的编码效应机理的基础上，设计基于稀疏表示的冗余3D字典，提出基于稀疏表示的编码效应消除方法；在同时考虑图像变换和实时渲染以及裸眼3DTV技术特点的基础上，建立基于非闭合区域最小化的高质量双目视图生成方法，改善3DTV的图像质量和立体效果；期望在获得3DTV的高质量双目视图和克服当前播放系统的带宽限制上取得实质性进展。

3DTV 是广播电视领域的一种新的播放模式，它可提供更加自然、生动的家庭视觉娱乐，其中存在的 编码效应和带宽问题已经成为实际应用的瓶颈问题。本项目按照资助项目计划书分析了降低图像质量和立体效果的编码效应机理，设计了基于稀疏表示的冗余3D字典，提出了基于稀疏表示的编码效应消除方法。在考虑图像变换和实时渲染以及裸眼 3DTV 技术特点的基础上，建立了基于非闭合区域最小化的高质量双目视图生成方法，改善了 3DTV 的图像质量和立体效果。通过视觉舒适度评估，增强了视觉舒适度，提出了基于自由能原则的感知率失真优化方法，提出了使用确定性区域检测和自适应视差调整的边缘保护立体匹配 方法，搭建了一个基于快速加权中值滤波算法的立体3D成像系统。