基于运动与深度感知的立体视觉舒适度研究

The depth cognition comfortable of stereo vision is throughout the entire three-dimensional information system. Stereo vision motion comfort is the main key problems and difficulties which must be solved. Perfect reproduction of the 3D scene in real-world with complex spatial and temporal dynamic motion relationship, is the goal that humans have been pursuing. Exploring the design methods of behavioral and objective experiments about stereovision comfortable motion features together with researching the incentives and scope of the stereo cognitive discomfort caused by motion, is the first critical scientific issue to be resolved. Dimensional cognitive process of the visual system is the joint processing by using two eyes simultaneously. Based on the principle of the human visual imaging system, to create a three-dimensional visual comfort motion model of the two eyes jointly working is the second critical scientific issue to be resolved. Stereo vision motion comfortable imaging is the ultimate goal of the three-dimensional display. Through the establishment of the eye tracking experiment platform and the stereo vision irritated platform of the stereo vision, to set up an about two eyes joint test and verification experimental system based on motion information feedback, is the third key scientific issue intended to be solved. The results of this research project are able to provide the theoretical basis and technical methods for solving the visual comfort issues caused in the three-dimensional display technology. Its achievements can promote the development of three-dimensional information technology significantly.

立体视觉认知舒适度问题贯穿于整个立体信息系统，是其中必须解决的关键问题和难点问题之一。完美再现真实世界的三维立体场景中复杂的时空动态运动关系，是人类一直追求的目标。探索符合立体视觉运动舒适特征的行为级实验方法，研究由运动引起的立体视觉深度认知不舒适的诱因和作用范围，是拟解决的第一个关键科学问题；视觉系统的立体认知过程是双眼并用的过程。依据人类视觉成像系统的原理，建立立体视觉舒适运动的人眼视觉模型，是拟解决的第二个关键科学问题；立体视觉运动舒适成像是立体显示的终极目标。通过建立立体视觉刺激显示平台和眼部跟踪平台，构建基于运动信息反馈的立体视觉舒适眼部实验系统，是拟解决的第三个关键科学问题。本项目的研究成果，能够为解决立体显示技术中立体视觉不舒适问题提供理论依据和技术方法，其成果对推动立体信息技术的发展具有重要意义。

本项目从图像质量评价、舒适度评估及应用等关键理论和技术开展了研究：(1)基于对图像失真原理的理解、分析，先后提出了高效的单失真、多失真及混合失真无参考图像质量评价算法；(2) 针对色调映射处理图像的处理机制及失真特性，从视觉感知处理过程模拟、图像多种属性分析等角度出发，提出多种高效的无参考色调映射图像评价算法；(3) 比较立体图像与平面图像的异同，基于双目视觉特性，结合立体图像失真特性，提出多种引入深度特征的无参考立体图像质量评价算法；(4) 比较立体图像质量与观看舒适度的异同，根据实际需要，搭建了主观测试平台（基于脑电、眼动等测试设备），设计主观实验方案，总结、分析、整理、推导实验结论；(5) 探索立体观看不舒适产生诱因，分析各个因素对舒适度的影响，建立客观舒适度感知模型；(6) 基于立体观看舒适度模型，首次提出立体字幕编辑工作，设计了自适应立体视频字幕位置判别算法。