【作者】 邵桂芳；

【导师】 李祖枢；

【作者基本信息】 重庆大学 ， 模式识别与智能系统， 2003， 硕士

【摘要】 虚拟环境下的实时自助照相系统的研制是为了满足体验经济的发展对数字应用系统的需求。该系统主要由人像的采集、提取和融合三部分组成，对前两部分的研究已经取得了较好的成果。本文首先从计算机视觉的理论体系和实际应用价值的角度上，分析了对低层次视觉要素的提取与检测的理论及其实际意义。分析了图像融合技术在基于PC的虚拟现实技术应用领域的地位，指出图像融合技术研究的意义；明确了进行本课题的必要性，提出了本课题的主要研究目标，确定了所需完成的研究任务。 鉴于目前对图像融合技术在娱乐领域的应用缺乏研究，尤其缺乏将虚拟现实技术、自助照相技术、机器视觉技术和图像融合技术综合运用的分析研究，因此我们从分析人类视觉特征入手，运用系统的观念综合平衡图像质量、系统成本等技术指标，在对虚拟实时自助照相系统具体技术指标深入分析的基础上，提出了基于对比度感知特征（Character of Apperceiving Contrast）的图像融合算法。这样，不仅能把信息融合技术（Information Fusion Technology，简称IFT）和人类视觉系统（Human Vision System，简称HVS）的研究成果引入虚拟照相系统中，使系统的“人性化”功能增强；同时，也对解决传统键控技术出现的“硬色键”问题，有一定的理论和实际的参考意义。摄像头受外界光学环境的影响，采集的人像经常有颜色失真现象，本文通过对HVS的颜色恒常性特征的分析，提出了基于HVS颜色恒常性的肤色校正算法，可以自动校正和补偿因外界影响而发生变化的人像肤色。在分析传统的图像质量评价方法及其缺陷的基础上，指出将HVS特性引入图像计测方法中的必要性；同时，对融合图像质量评价现状进行了深入的研究，鉴于本课题研究的特殊性，建立了一套适合娱乐照相领域融合图像的质量评价方法，并给出实验分析。总结本文的工作，文中通过理论分析、实际应用和理论上升加仿真模拟三个阶段的工作对图像融合在虚拟环境自助照相系统中的应用进行了深入的探索。各种实验结果证明了文中相关理论的合理性和正确性，同时验证了文中所提算法是可靠的，具有较强的鲁棒性，并表明了图像融合技术在实际应用中有着广阔的应用空间。更多还原

【Abstract】 The Self-Control Photo System can meet the higher requirement for Digital Application systems in experience economic booming period. It consists of Image Acquisition Unit, Image Extraction and Image Fusion Unit. The research on the first two units presents satisfying results.At the beginning, this paper discusses the Low-level Vision Elements Extraction and Detection theory, especially in Computer Vision systems and its practical meaning. Then the roles and prospects of Image Fusion in Virtual-Reality application based on PC are analyzed. Following the possibility and necessary analysis, the major research missions and expected targets are constructed. According to the deficiency of Image Fusion applications in entertainment equipments, such as no special research that combines Virtual Reality, Self-control Photographing, Computer Vision and Image Fusion technologies together, we propose an Image Fusion algorithm base on Character of Apperceiving Contrast, which take system elements into account, e.g. integrity image quality balance and costs. Based on the entire analysis of Self-Control Virtual Real-time Photographing, the algorithm not only introduces the Information Fusion theory and Human Vision System to Virtual Photographing system, but also resolves the hard color-key problem caused by traditional key control method. It reveals humanistic function and good practical value.The image probably occurs color distortion that is influenced by optical equipment. In this paper, the HVS color invariance characters is construed, an complexion rectify algorithm based that is proposed, which can automatically adjust and compensate the image quality according to the external environment influence.After summarized traditional evaluation methods and research on the Image Fusion trends, it is obvious to adopt HVS method to achieve good results. An image quality evaluation method is proposed, which is entertainment-oriented, appended with experiment results.This paper studies the Image-Fusion in Self-control Virtual Photographing through three periods working, which are theory analysis, practical implantation and theory breakthrough with simulation. The experiment results prove that the theory and the algorithms are correct; also it displays good system stability, robust quality. All of above reveals the great application potential of Image Fusion technology.更多还原