基于人类多元感知特性的全景图像质量客观评价理论与方法

The global development of panoramic technology provides us an opportunity to enhance the international competitiveness of our domestic panoramic technology. Up to now, there is no thorough method to evaluate the panoramic image quality, and no corresponding industry standard and evaluation system at domestic and overseas. Therefore, it has become an urgent problem to break through the panoramic image quality assessment technology. This project intends to research this topic based on human multi-perception characteristics, namely human visual physiological perception characteristic, human visual psychological perception characteristic and human aesthetic perception characteristic. Firstly, based on human visual physiological and visual psychological perception characteristics, this project studies human visual multi-perception characteristic model to improve the efficiency of the panoramic image quality assessment method. Secondly, the aesthetic perception characteristic of panoramic image is studied based on the strong demand of human spiritual aesthetic perception, and the visualized mathematical model of human aesthetic perception characteristic is proposed to improve the completeness of the panoramic image quality assessment method. Finally, by integrating human physiological, psychological and aesthetic multi-perception characteristics, the evaluation indexes of the physiological quality and spiritual aesthetic perception of panoramic image are studied, and the universal objective panoramic image quality assessment theory and method are then established. The research results of this project will promote the further development and implementation of our domestic panoramic image technology, which meets the country's major development strategy needs, and people's colorful work and life needs.

全景技术的全球化发展为提升我国全景技术的国际竞争力提供了良好机遇。目前，国内外尚未有完善的全景图像质量评价方法，也缺乏相应的行业标准与评价体系。因此，突破全景图像质量评价技术已成为亟待解决的问题。本项目拟从人类多元感知特性入手，即视觉生理感知特性、视觉心理感知特性以及美学感知特性，展开深入研究。首先基于人类视觉生理与视觉心理感知特性，深入研究视觉多感知特性模型，提升全景图像质量评价的高效性；其次，拟结合人类对精神美感的强烈需求，探究全景图像美学感知特性，提出可视化的美学感知特性数学模型，提升全景图像质量评价的完备性；最后，综合人类视觉生理、视觉心理与美学感知多元感知特性模型，研究全景图像生理质感与精神美感评价指标，建立普适性的全景图像质量客观评价理论与方法。本项目研究成果将推动我国全景图像技术的进一步发展与落地，满足国家重大发展战略需求，满足人们丰富多彩的工作与生活需求。

全景图像是新一代信息获取、传播与显示的研究核心，在军事、教育、娱乐等众多领域都有广阔的应用前景。由于全景图像视野广、信息量大，有时需要手、眼、头等多感官协作去理解感知其内容，传统图像质量评价方法已不能有效度量全景图像质量。针对上述问题，本项目通过分析人类多元感知特性，有针对性地研究了人类视觉生理、心理、精神美学感知特性，提出了全景图像质量客观评价方法。主要研究内容包括：（1）深入分析了全景图像视觉生理、心理感知特点，结合多元感知特性关联关系，提出了基于视觉生理和心理感知特征融合的、基于多尺度局部和全局视觉感知失真的、基于超像素视觉显著性特性的、基于视窗分割与自然统计特性的全景图像质量客观评价方法；（2）结合全景图像美学构图特点，剖析了高层美学感知机制与建模方法，研究了全景图像美学感知质量评价方法；（3）基于人类视觉和美学特性关系，有效融合了低级质量以及高级美学感知特性，提出了多元感知融合的全景图像质量评价方法。本项目的研究从理论和实验上揭示了全景图像视觉、生理和心理感知特性，从不同角度提出了多种有效的质量评价算法，达到了项目预期目标。本项目研究成果可为虚拟现实图像的设计与优化、性能评价等提供强有力的工具，具有重要的理论意义和实用价值