摘要
为检测人脸考勤机在采集人脸照片时是否存在过曝光问题,提出一种基于图像亮度信息、颜色信息和空间区域特征的图像过曝光检测算法。首先,基于sigmoid函数设计过曝光置信度来衡量像素点过曝光程度,然后进行参数调优以确定最优模型。在选取的实验图像上的实验结果表明,LCS算法检测出的过曝光区域呈现整体性和均匀性,更加符合人眼视觉习惯,在检测效果上优于简单阈值法和基于亮度和颜色特征的LC算法。
In order to detect the existence of over-exposure problem when face recognition equipment is collecting face photos,proposes an image over-exposure detection algorithm based on image lightness information,color information and spatial region features.Firstly,designs overexposure confidence to measure the degree of over-exposure based on sigmoid function and then tuned parameters to determine the optimal model.The results on the selected experimental images show that the over-exposure regions detected by the LCS algorithm are integral and uniform,which is more in line with human visual habits and superior to the simple threshold method and the LC algorithm.
作者
周明明
龚敏
ZHOU Ming-ming;GONG Min(College of Computer Science,Sichuan University,Chengdu 610065;School of Mathematic and Statistics,Southwest University,Chongqing 262500)
出处
《现代计算机》
2018年第6期46-50,共5页
Modern Computer
关键词
过曝光检测
人脸图像
区域
指标设计
Over-Exposure Detect
Face Image
Region
Indicator Design
