中国科学院沈阳自动化研究所机构知识库
Advanced  
SIA OpenIR  > 机器人学研究室  > 学位论文
题名: 基于物理成像机理的图像光照建模及处理方法研究
其他题名: Physical and Imaging based Illumination Modelling and Processing for Images
作者: 田建东
导师: 唐延东
分类号: TP391.4
关键词: 光照建模 ; 成像计算 ; 反射光谱计算 ; 阴影建模与处理 ; 光照转换
索取号: TP391.4/T56/2011
学位专业: 模式识别与智能系统
学位类别: 博士
答辩日期: 2011-06-01
授予单位: 中国科学院沈阳自动化研究所
学位授予地点: 中国科学院沈阳自动化研究所
作者部门: 机器人学研究室
中文摘要: 作为自然界中普遍存在的物理现象,室外场景的光照变化给计算机视觉算法的鲁棒性带来诸多不利影响。它一直是计算机视觉以及相关学科的重要研究内容,但一直未得到较好的解决。不同于以往仅对图像数据做数学处理或机器学习的方法,本论文将从研究物理成像机理的角度出发对图像中的光照现象进行建模与分析。物理成像特性包含光谱成像特性和几何成像特性两个方面。其中光谱成像特性主要依赖于光源的光谱特性、物体的反射特性及相机的成像特性,是图像中物体颜色和亮度的主要决定因素,也是我们的主要研究内容。本论文具体的研究内容包括室外光谱辐照度计算、阴影的建模与处理、相机建模、反射光谱计算及光照变换。在分析光谱成像特性的基础上,我们重点研究了阴影的建模与处理算法。我们对所提出的模型和算法均进行了实验验证,并将这些光照处理算法成功应用于目标识别与跟踪。此外,本论文的研究成果(包括方法、模型、算法)亦可应用于如计算机图像学、图像渲染、虚拟现实、场景分析等其它相关领域的光照处理之中。 本论文的创新工作主要包括: 1. 论文的创新点首先体现在研究思路和方法上。目前,计算机视觉中光照处理的主流方法仍是把图像作为数据进行数学处理或机器学习;基于物理的方法也只是针对特定现象如雾等进行处理,并无整体框架。本论文对物理成像所涉及的光照、反射、相机响应均进行了分析与建模,并紧密结合图像中的光照处理,提出了相应的算法,形成了较为系统的理论和算法框架。 2.      针对室外阴影,利用其太阳直射光被遮挡而只受天空散射光照射这一几何特性,应用太阳光和天空光的光谱特性和相机成像机理,推导了阴影区域和对应的非阴影背景之间的三色衰减模型,并在此基础上提出了阴影检测算法。 3.      针对室外光源(太阳光和天空光)的多变性且缺乏有效的分析计算工具这一现状,利用外太空辐射光谱经过地球大气层时的吸收、散射规律,结合人眼和相机的响应特性,在计算机视觉领域首次提出了太阳光和天空光地表光谱辐照度的可行性计算方法。 4.      利用太阳光和天空光在不同天顶角度的光谱辐射特性,推导了阴影内部和外部的线性模型,并证明了该模型的参数只与光源的光谱有关,而与物体的反射光谱无关。在此模型基础上提出了本征图像获取、阴影判定、阴影去除的算法。 5.      提出了不依赖于具体设备,不同光源下最优的相机光谱响应曲线计算方法;提出了反射光谱曲线重建的最优基函数计算方法。在此基础上,给出了图像的光照转换方法。
英文摘要: As one common physical phenomena, the variation of illumination in outdoor scenes causes many problems to the robustness of computer vision algorithms. Though it is always an important research topic in computer vision and related scientific branches, it has not solved well yet. Different with previous methods that only do mathematical or learning-based processing on image data, in this dissertation, illumination phenomenon in images are modelled and studied based on analyzing physical and imaging mechanism.  Physical imaging properties contain those of spectrum and geometry. Spectral properties rely on spectrum of light sources, surface reflectance, and camera imagery. Spectral properties determine the color and intensity of an image and are the main research content of this dissertation. In detail, it includes calculating spectral power distribution (SPD) of sunlight and skylight, modelling and processing shadows, modelling image formation and recovering spectral reflectance, and converting an image under one illumination to another. The modelling and processing methods for shadows are principally studied based on analyzing the spectral imaging properties. With experimental results and comparing with other models and algorithms, we validate our proposed models and algorithms in the dissertation. We also successfully applied our illumination processing methods to objects recognition and targets tracking. The results of the experiments and the applications show the good performance of our proposed models and algorithms. In addition, our proposed illumination models and processing algorithms for computer vision can be also applied in other related research fields such as computer graphics, image rendering, virtual reality, and scene understanding. The original work in this dissertation mainly includes: 1.      Our originality is firstly reflected by research methodology. So far, the mainstream of illumination processing methods mainly rely on mathematical analysis for image data and machine learning for image features. Some physical-based methods just focus on some specific scopes, e.g., foggy image processing. In this dissertation, illumination, reflectance, and camera spectral response functions are analyzed and new models are proposed. Based on these models, we developed some illumination processing algorithms. Overall, our proposed models and algorithms form a new physical-based framework about illumination processing in computer vision. 2.      Based on the fact that a shadow is illuminated by skylight while is shaded from direct sunlight and combining the properties of imaging and the SPDs of sunlight and skylight, we proposed the Tricolor Attenuation Model (TAM) for shadows. According to the model, two shadow detection algorithms are proposed. 3.      Based on the rules of absorption and scattering when solar irradiance pass through the earth’s atmosphere and combining the characteristics of human eye and cameras, we firstly proposed the computer vision feasible calculation method for direct sunlight and diffuse skylight. 4. Based on the properties of sunlight and skylight at different sun angles, we deduced the linear model between a shadow and its non-shadow background. We also proved that the parameters of the linear model only depend on the spectrum of light sources rather than object reflectance. We further proposed algorithms for intrinsic-images deriving, shadow verification, and shadow removal. 5. We proposed the calculation method for optimal spectral response functions of cameras that are device-independent and illumination-based. We also proposed the calculation method for optimal basis functions in reflectance recovery. Based on them, we designed a method for lighting conversion for a color image.
语种: 中文
产权排序: 1
内容类型: 学位论文
URI标识: http://ir.sia.cn/handle/173321/9388
Appears in Collections:机器人学研究室_学位论文

Files in This Item:
File Name/ File Size Content Type Version Access License
基于物理成像机理的图像光照建模及处理方法研究.pdf(12157KB)----限制开放 联系获取全文

Recommended Citation:
田建东.基于物理成像机理的图像光照建模及处理方法研究.[博士学位论文].中国科学院沈阳自动化研究所.2011
Service
Recommend this item
Sava as my favorate item
Show this item's statistics
Export Endnote File
Google Scholar
Similar articles in Google Scholar
[田建东]'s Articles
CSDL cross search
Similar articles in CSDL Cross Search
[田建东]‘s Articles
Related Copyright Policies
Null
Social Bookmarking
Add to CiteULike Add to Connotea Add to Del.icio.us Add to Digg Add to Reddit
所有评论 (0)
暂无评论
 
评注功能仅针对注册用户开放,请您登录
您对该条目有什么异议,请填写以下表单,管理员会尽快联系您。
内 容:
Email:  *
单位:
验证码:   刷新
您在IR的使用过程中有什么好的想法或者建议可以反馈给我们。
标 题:
 *
内 容:
Email:  *
验证码:   刷新

Items in IR are protected by copyright, with all rights reserved, unless otherwise indicated.

 

 

Valid XHTML 1.0!
Copyright © 2007-2016  中国科学院沈阳自动化研究所 - Feedback
Powered by CSpace