摘要
在传统的基于小波变换的人脸识别方法的基础上,加入稀疏表示的方法对人脸识别进行研究,进一步提高人脸识别率。小波变换把人脸图像分解为一幅低频人脸图像和三幅高频人脸图像,低频人脸图像代表人脸图像的全局(整体)信息,高频人脸图像代表人脸图像的纹理和边缘等细节信息。低频人脸图像在人脸识别中起到关键性作用,用正交投影的方法对低频人脸图进行识别得到的低频人脸图像分类隶属度。高频人脸图像在人脸识别中同样存在不可忽略的作用,用基于领域能量的方法把三幅高频人脸图像融合为一幅高频融合人脸图像,然后用稀疏表示的方法对融合图像进行识别得到高频人脸图像分类隶属度。最后把高、低频分类隶属度融合确定人脸图像所属类别,与传统人脸识别方法相比,进一步提高了人脸识别率。
The tranditional face recognition algorithm based on wavelet transformation, using sparse representation to recognize the ignored high-frequency sub-band images and providing higher recognition rates of the face recognition. The Wavelet Transformation transform the face images into a low-frequency sub-band images and three high-frequency sub-band images, as low- frequency sub-band images representing the face image globally （Integrally） information, it plays a key role in face recognition, then calculating the classification membership degree by recognize the low-frequency sub-band images; High-frequency sub-band images with image information including horizontal, vertical and diagonal image , presents the face texture and edge details, also plays a non-negligible role in face recognition. Decomposed high-frequency sub-band image＇s fusion based on the field of energy can export a high- frequency sub-band fused image which recognized by the sparse representation method, and the classification membership degree used in high-frequency part of the face recognition can be got. Finally, a dynamic weighted fusion method can fusion the two classification memberships and obtain the final classification of membership degree which improves face recognition rate and used in the final face＇s classification and identification.
出处
《中国科技信息》
2014年第8期155-158,共4页
China Science and Technology Information
基金
黑龙江省教育厅科学技术研究项目(NO:12533054)
关键词
小波变换
稀疏表示
图像融合
人脸识别
wavelet transformation
sparse representation
face recognition
image fusion
