基于分割融合算法的WCE过曝光图像修正研究  

Research on Multi-exposure Image of WCE based on Segmentation and Fusion Algorithm

作　　者：魏晓雍 王孟 曹阳[1,2] 张大斌[1] 刘杰民[3] 倪茂美[3] WEI Xiaoyong;WANG Meng;CAO Yang;ZHANG Dabin;LIU Jiemin;NI Maomei(Guizhou University,Guiyang Guizhou 550025,China;Engineering Training Center,Guizhou University,Guiyang Guizhou 550025,China;Guizhou Provincial People’s Hospital,Guiyang Guizhou 550001,China)

机构地区：[1]贵州大学,贵州贵阳550025 [2]贵州大学工程实训中心,贵州贵阳550025 [3]贵州省人民医院,贵州贵阳550001

出　　处：《通信技术》2021年第5期1095-1102,共8页Communications Technology

摘　　要：当无线胶囊内镜(Wireless Capsule Endoscopy,WCE)工作时,进入摄像头的光源超过图像传感器能接收的范围,会导致拍摄图像产生过曝光现象,导致图像中存在颜色、细节信息缺失,从而影响医生或机器的诊断。因此,提出了一种基于分割融合算法的WCE图像过曝光区域修正方法。首先,对肠部息肉和溃疡局部曝光进行图像增强,降低普通区对过曝光区域的颜色影响;其次,提取图像的亮度和饱和度分量,分别采用模型1和模型2分割得到强、弱爆光区域分量,并在弱爆光区经拉普拉斯锐化处理后突出边缘信息;最后,将两分量融合得到过曝光区域掩膜,采用邻域镜像替换法填充修复得到修正后的图像。通过软件Matlab 2016进行仿真实验,结果表明,图像过曝光区域修正前后对比,像素点个数平均下降了97.8%,修正后得到了更加符合人眼所感知的图像,明显改善了图像纹理、颜色等,验证了所提算法对过曝光区域纹理和颜色增强的有效性。When the Wireless Capsule Endoscope(WCE)is working,the light source entering the camera exceeds the range that the image sensor can receive,which will cause multi-exposure in the captured image,and the lack of color and detail information in the image will affect the doctor or machine diagnosis.This paper proposes a method of multi-exposure area correction of WCE image based on segmentation fusion algorithm.Firstly,it performs image enhancement processing on the multi-exposure of typical stomach and intestinal polyps and ulcers,the purpose is to reduce the effect of common areas on the color of multiexposure.Then,by extracting the brightness and saturation components of the image,models 1 and 2 are used to obtain the strong and weak multi-exposure components,and the weak exposure area is processed by Laplace change to highlight the edge information.Finally,the two components are merged to obtain the multi-exposure mask map,and the neighborhood mirror replacement method is used to fill and repair the multi-exposure correction futures.The simulation experiment is carried out through the software MATLAB 2016,and the results indicate that the number of pixels is decreased by 97.8%,the effectiveness of the proposed algorithm for texture and color enhancement in the multi-exposure areas is verified.

关 键 词：分割融合算法 过曝光图像 图像增强 拉普拉斯锐化 镜像替换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]