摘要
文中基于二进小波变换,将图像的颜色、形状以及纹理3种不同的特征提取方法相互融合,对人脸图像的辨别性特征进行提取与检索。采用3种不同的方法来表达图像的特征信息,并在两种不同的标准人脸数据库中执行实验操作。由结果可知,所获得的特征蕴含了很多内容,因此整体地对图像内容进行了各种补充,该方法对人脸图像中存在的光照变化、姿态、表情、位置、时间以及遮挡物等因素有很高的鲁棒性。
Based on dyadic wavelet transform,this paper uses three different feature extraction methods of image color,shape and texture to extract and retrieve the discriminative features of face image.Three different methods are used to express the image feature information,and experimental operations are performed in two different standard face databases.It can be seen from the results that the obtained features contain a lot of content,so the image content is supplemented as a whole.This method has high robustness to the illumination change,posture,expression,position,time,occlusion and other factors in the face image.
作者
阿斯古丽·艾合麦提
吐尔洪江·阿布都克力木
赵思温
Asigul Ahmat;Turhunjan Abdukirim;ZHAO Siwen(School of Mathematics and Data Science,Changji University,Changji 831100,China;School of Mathematical Science,Xinjiang Normal University,Urumqi 830017,China)
出处
《电子设计工程》
2022年第15期185-188,193,共5页
Electronic Design Engineering
关键词
二进小波变换
边缘检测
LBP算子
特征提取
检索
dyadic wavelet transform
edge detection
LBP operator
feature extraction
retrieval