面向用户视觉体验的立体全景视频质量评价及增强方法研究

全景视频作为一种新型的虚拟现实视频，可提供沉浸式、临场式的观看体验，是新一代信息技术的发展方向和趋势。本项目针对立体全景视频在质量评价和增强方面的技术瓶颈，将从功能模拟的角度解决以下三个关键科学问题：1）立体全景视频在不同空间的视觉表现形式是什么？2）面向显示的立体全景视频质量评价模型是什么？3）面向用户视觉体验的立体全景视频质量增强技术是什么？本项目紧紧围绕上述关键科学问题，拟分别在立体全景视频数据库构建、视觉体验质量计算模型、视频质量评价和视觉体验质量增强四个层面展开研究。拟通过开展主观实验构建带有质量标签的立体全景视频数据库，形成多模态融合的视觉体验质量超完备表示的神经元字典库，并基于此提出视觉质量评价及增强方法，以达到优化视觉体验质量的目的。通过本项目的研究，力争在立体全景视频质量评价及优化问题上实现核心理论和关键方法的突破，为虚拟现实技术的研究提供理论基础。

Panoramic video is a new type of virtual reality video in providing an immersive and on-the-spot viewing experience for users, which is expected to be the development direction and trend of the new generation of information technology. To solve the technique bottleneck of stereoscopic panoramic video in quality assessment and enhancement, this project aims to solve the following three fundamental scientific issues: 1) What are the manifestations of stereoscopic panoramic vision in different spaces? 2) What are the quality assessment models of stereoscopic panoramic video for display? 3) What are the enhancement models of stereoscopic panoramic video for user’s quality of experience? Focused on the above frontier scientific issues, we carry out the research work on the following four aspects: establishing stereoscopic panoramic video database, modeling quality of experience, video quality assessment and video quality enhancement. Specifically, by constructing stereoscopic panoramic video database with multiple quality labels and analyzing the influence on quality of experience, we establish a multi-modal neural dictionary for overcomplete representation of quality of experience, and use this model for further quality evaluation and quality enhancement to optimize the quality of visual experience. Through this project, it will form cognitive theories and calculation models for stereoscopic panoramic vision quality assessment and enhancement, and provide further theoretical support for the research of virtual reality technology.