改进导引滤波器的人脸光照预处理方法  

Face light preprocessing method based on improved guided filter

作　　者：吴海英[1] 张光辉 桂预风[1] WU Haiying;ZHANG Guanghui;GUI Yufeng(School of Science,Wuhan University of Technology,Wuhan 430070,China)

机构地区：[1]武汉理工大学理学院,武汉430070

出　　处：《中国科技论文》2018年第2期230-234,共5页China Sciencepaper

基　　金：教育部人文社科基金资助项目(12YJAZH022);湖北省统计科研计划重点项目(HB131-25);湖北省商务厅科研资助项目(HBSW-2014-01)

摘　　要：针对光照对人脸识别的影响,对传统的导引图像滤波器进行改进,在进行滤波之前,先利用具有边缘检测特性的全局梯度信息的权值模型计算原图像各像素的权模,并将其作为滤波的自适应参数,使得在滤波过程中滤波器能够依据图像的具体内容而自适应地改变滤波参数。将改进方法与基于LoG算子的方法进行比较,结果表明,改进方法可以有效地改善对自熵图像的识别效果,提高人脸图像的识别率。To eliminate the influence of illumination on face recognition,the traditional guided image filter is improved in this paper.Before filtering,the weighting model of global gradient information with edge detection characteristics is used to calculate the weight of each pixel in the original image,and the weight is used as the adaptive filter parameters,which makes the filter adaptively adjust the filter parameters according to the specific content of images in the filtering process.Compared with the improved method based on operator of LoG,our method can effectively enhance the recognition effect of self-entropy image and improve the final face recognition rate.

关 键 词：自熵图像法 导引图像滤波器 梯度 自适应函数 滤波参数 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]