利用MRA-框架的人脸识别方法的研究

Research of Face Recognition Method Use of MRA-Framework

摘要: 图像拼接中出现的运动目标可能使拼接出现不能正常拼接或者拼接出多重影像的现象。本文提出一种图像拼接的运动目标检传统的PCA方法，人脸图像加噪后，人脸识别率会明显下降。本文针对这种情况，分别利用正交小波 + PCA和小波框架 + PCA方法进行了研究.首先对人脸图像进行加噪处理，然后对图像进行正交小波和小波框架分解，进而对分解后的子图分别利用PCA方法进行降维和特征提取，最后用三阶近邻法作为分类器进行分类识别。通过ORL人脸数据库的验证，结果证明了本文方法的有效性，很好的提高了加噪情况下人脸图像的识别率。

Abstract: After adding noise, face recognition rate of the traditional PCA method will be significantly lowered. This paper will use methods of orthogonal wavelet + PCA and wavelet frame + PCA to study it respectively. First, we add noise to deal with the image, then decompose the image under the use of orthogonal wavelet and wavelet frame; next, for the subgraph that has decomposed we will reduce the dimensionality and feature extraction using PCA method respectively; finally, we use third-order nearest neighbor as the classifier to classify and identify it. Through the test and veri-fication of the ORL face database, it shows the effectiveness of this method, which is a good way to improve the recognition rate of face image under the condition of adding noise.