多尺度Retinex算法在简牍图像增强中的应用  

Application of Multiscale Retinex in JianDu Image Enhancement

作　　者：汪刘艳[1] 王绪本[1] 

机构地区：[1]四川成都理工大学信息工程学院,成都610059

出　　处：《计算机科学》2009年第3期288-290,共3页Computer Science

基　　金：国家"十五"科技攻关项目"文物保护关键技术研究"(2004BA810B)资助

摘　　要：简牍图像增强是简牍文字修复和文字信息提取最为关键的一个步骤。以长沙简牍博物馆收藏的简牍文物为实物模型,以该馆提供的简牍数字图像为研究对象,提出了基于多尺度Retinex(MSR)算法的简牍图像增强方法,论述了本算法的原理及实现方法。采用MSR算法来消除光照对简牍图像的退化影响,以达到增强图像暗区细节信息的目的;同时针对MSR输出图像偏暗,提出了直方图拉伸算法。试验结果表明:本算法能够提高简牍图像对比度的同时,能显著地增强图像暗区的细节信息并且使简牍图像具有较好、较高的色彩保真度。本方法克服了目前常规的图像增强方法的不足,能够满足简牍考古人员的要求。Enhancement of the JianDu Image is a key procedure in character reconstruction and text extraction. We made the JianDu relics from Changsha Jiandu museum as a model and taken the Jiard）u images as our research objects. An enhancement algorithm of JianDu image based on multiscate Retinex was presented. Based on the theory of Retinex, the multi-scale Retinex（MSR） algorithm was employed to remove the luminance degradation 05 the Jianl）u image. The information of hidden area of the JianDu image was obviously enhanced. For the intensity of the MSR processing output is weak,one normal histogram stretching method was deployed to improve the output brightness. The results of experiment show that MSR can improve effectively not only the contrast but also can observably enhance the information of hidden area in the JianDu image and provided better tonal rendition. The algorithm can overcome the lack of enhancement of traditional enhancement methods and satisfy the archaeologist＇s demand.

关 键 词：RETINEX理论 多尺度RETINEX算法 直方图拉伸 简牍图像 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]