计算机视觉中自然光照建模及其恒常性计算

项目名称： 计算机视觉中自然光照建模及其恒常性计算

项目编号： No.61473280

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 田建东

作者单位： 中国科学院沈阳自动化研究所

项目金额： 80万元

中文摘要： 复杂多变的光照环境给计算机视觉算法及应用（如特征提取，目标分割与识别、测量）带来诸多问题，降低了其算法的鲁棒性及环境自适应性。寻求该问题的有效解决方案一直是计算机视觉及相关学科的重要研究内容。不同于目前基于数学统计或机器学习的研究方法（单纯数据驱动研究方式），我们将从基本的大气物理、物理光学原理及物理成像机理出发（物理原理和数据联合驱动方式），研究图像中光照变化问题。从光照物理成像特性分析的角度去研究和处理问题，建立新的模型、提出新的观点、取得具有原创性的理论成果。主要研究内容包括：建立计算机视觉中可计算的室外光照模型；建立高精度反射光谱恢复及成像仿真计算模型；提出符合物理成像机理且具有环境自适应能力的图像光照处理算法（本征图像分解算法、图像光照转换及光照恒常算法）。从目前的研究现状看，这将是一个全新的研究思路。我们已取得了初步研究成果，为该课题研究内容的实施打下了良好的基础。

中文关键词： 计算机视觉；视觉感知；目标识别

英文摘要： The complicated and changefully illumination environments bring many problems to computer vision and its application, such as feature extraction, object segmentation, recognition and measurement. They degrade the performance of the algorithms in computer vision and the adaptiveness of these algorithms to the environments. It is one of the most important research topics how to make the algorithms of image processing robust to the changeful illumination environments. Different from the state-of-the-art algorithms with mathematics and machine learning (only with data driven), we will develop the research on the illumination problems in image processing based on theories of atmospheric physics, physical optics, imaging mechanism and from the view of characteristic analysis of physical imaging. The aim of our proposal is to set up new models, to propose novel viewpoints and algorithms for effective processing illuminant problems in image processing, and to establish new theory in this research topic. Our research content includes the calculable illumination model for computer vision, reflectance spectroscopy calculation of an image, algorithm of intrinsic image, color constancy algorithm and illumination converting of an image. It is a completely new research way for these problems. We have made principium progress in this topic research, which provides a strong background for our research in this proposal.

英文关键词： computer vision;visual perception;object detection;illumination modelling;intrinsic image