分类号密级UDC注1学位论文基于Retinex理论的图像增强算法研究(题名和副题名)陈雾(作者姓名)指导教师姓名任明武申请学位级别硕士专业名称模式识别与智能系统论文提交日期论文答辩日期学位授予单位和日期答辩委员会主席评阅人年月日注1:注明《国际十进分类法UDC》的类号。声明本学位论文是我在导师的指导下取得的研究成果,尽我所知,在本学位论文中,除了加以标注和致谢的部分外,不包含其他人已经发表或公布过的研究成果,也不包含我为获得任何教育机构的学位或学历而使用过的材料。与我一同工作的同事对本学位论文做出的贡献均已在论文中作了明确的说明。研究生签名:年月日学位论文使用授权声明南京理工大学有权保存本学位论文的电子和纸质文档,可以借阅或上网公布本学位论文的部分或全部内容,可以向有关部门或机构送交并授权其保存、借阅或上网公布本学位论文的部分或全部内容。对于保密论文,按保密的有关规定和程序处理。研究生签名:年月日硕士论文基于Retinex理论的图像增强算法研究I摘要Retinex是一种建立在科学实验和科学分析基础上的图像增强理论,其实质就是从图像中抛开照射分量的影响来获得物体的反射分量,即获得物体本来的面貌。与其它图像增强方法相比,Retinex算法具有锐化、颜色恒常性、动态范围压缩大、色彩保真度高等特点。故本文对该理论进行了重点研究。本文分析了几种常规的图像增强方法的优缺点;在颜色恒常性理论的基础上研究了全局和局部Retinex算法;对全局Retinex算法引入了两种改进方法;详细地分析了局部Retinex算法和同态滤波之间的相关性;并对在局部Retinex算法中使用的高斯模板的大小和高斯函数的方差之间的关系进行了研究;实验过程合理地分析和使用了几个有效的图像增强的评价指标。最后,本文在分析全局和局部Retinex算法的基础上,提出了一种将两者联合的Retinex方法。通过实验分析,该联合的Retienx算法在对比度和熵方面基本达到了局部Retinex算法的水平,且在亮度方面更是优于局部Retinex算法。关键词Retinex,同态滤波,颜色恒常性,动态范围压缩,信息熵硕士论文基于Retinex理论的图像增强算法研究IIAbstractTheRetinexisonekindofimageenhancementtheoryanditeatablishedatthescientificexperimentandthescientificanalysis.Itessenceisto‘skinning’theilluminationintheoriginalimageandtoyieldthereflectanceimage,thatis,theoriginalappearanceoftheobject.Compareswithconventionalenhancementmethods,theRetinexhasshownitselftobeaveryversatileautomaticmethodandcansimultaneouslyprovidesharpening,colorconstancy,dynamicrangecompressionandcolorrendition.Thisthesisfirstdiscussedseveralbasicimageenhancementmethods,andanalyzedtheshortcomingsofthem.ThentheglobalandlocalRetinexalgorithmswereinvestigatedindetail.ThisthesismadetwomodificationstotheglobalRetinexalgorithmtogetgoodresults.ItdiscussedtherelationbetweenthelocalRetinexalgorithmandhomomorphicfiltering.AfterwardsitpointedouttheGaussianfilterradiusshouldbethreetimesofthestandarddeviation.Duringtheexperiment,severaleffectiveparametersforimageenhancementappraisalhavebeenanalyzed.Finally,thisthesisproposedacombinedRetinexmethodintegratedtheglobalandlocalaspects.TheexperimenthadproventhecombinedmethodproposedinthisthesiscanreachthesamelevelwiththelocalRetinexatcontrastandtheentropyaspect,butcanbemorebrighterthanthelocalRetinex.Keywords:Retinex,HomomorphicFiltering,ColorConstancy,DynamicRangeCompression,Entropy硕士论文基于Retinex理论的图像增强算法研究III目录摘要...........................................................................................................................IAbstract.........................................................................................................................II目录........................................................................................................................III图表目录.........................................................................................................................V1绪论................................................................................................................................11.1问题的提出.........................................................................................................11.2研究现状及问题.................................................................................................21.3本文的研究工作.................................................................................................21.4本文各章内容安排.............................................................................................32图像增强方法概述........................................................................................................42.1线性变换和分段线性变换.................................................................................42.1.1线性变换.................................................................................................42.1.2分段线性变换.........................................................................................52.2图像边缘增强.....................................................................................................62.3直方图均衡化.....................................................................................................62.4直方图规定化.....................................................................................................92.5同态滤波.............................................................................................................93基于全局特征的Retinex图像增强............................................................................123.1颜色恒常性与Retinex理论.............................................................................123.1.1颜色恒常性............................................................................................123.1.2Land的Retinex理论............................................................................123.1.3Retinex理论的发展...............................................................................123.2基于全局特征的Retinex算法分析.................................................................133.2.1灰度图像增强........................................................................................133.2.2彩色图像增强........................................................................................163.3算法流程...............................................................