MATLAB下人脸图像的PCA重构matlab代码(共5页).doc

《MATLAB下人脸图像的PCA重构matlab代码(共5页).doc》由会员分享，可在线阅读，更多相关《MATLAB下人脸图像的PCA重构matlab代码(共5页).doc（5页珍藏版）》请在淘文阁 - 分享文档赚钱的网站上搜索。

1、精选优质文档-倾情为你奉上MATLAB下人脸图像的PCA重构matlab代码主函数clc; clear all;newimage=double(zeros(112*6,92*5); M,sort_V,avg_face = averageface( );%返回去均值的矩阵和协方差，以及排列好的特征向量 k=1; new_sort_V =sort_V(1:10304,1:30);%取特征向量的前30个，并进行显示 for i=1:112:5*112+1 for j=1:92:4*92+1 newimage(i:i+111,j:j+91)=reshape(new_sort_V(:,k),112,92

2、); k=k+1; end end minum=min(min(newimage);%对特征向量进行标定 maxum=max(max(newimage); d=maxum-minum; newimage(:,:)=newimage(:,:)-minum; newimage(:,:)=newimage(:,:)/d; newimage(:,:)=newimage(:,:)*255; newimage=uint8(newimage); subplot(1,3,1); title(特征向量图); imshow(newimage);%显示前30个特征向量 avg_face=uint8(avg_face

3、); subplot(1,3,2) title(平均脸); imshow(avg_face); subplot(1,3,3) tyxs,re_newimage = touyingxishu( sort_V ,avg_face); imshow(re_newimage); 子函数function =img2txt()maindir = D:Image; subdir = dir( maindir ); % 先确定子文件夹 m=1;k=1;for i = 1 : length( subdir ) if( isequal( subdir( i ).name, . ) | isequal( subdi

4、r( i ).name, . ) | subdir( i ).isdir ) % 如果不是目录跳过 continue; end subdirpath = fullfile( maindir, subdir( i ).name, *.pgm ); images = dir( subdirpath ); % 在这个子文件夹下找后缀为pgm的文件 % 遍历每张图片 for j = 1 : length( images ) imagepath = fullfile( maindir, subdir( i ).name, images( j ).name ); imgdata = imread( ima

5、gepath ); % 这里进行你的读取操作 row, col=size(imgdata); %将数据写入txt中 fid = fopen(C:UsersAdministratorDesktop编程实验材料att_faces,s,int2str(m),int2str(k),.txt,w); %新建txt文件存储图像的坐标和灰度值 for I=1:row for J=1:col fprintf(fid,%03dt,imgdata(I,J); end fprintf(fid,n) ; end fclose(fid); k=k+1; end m=m+1; k=1;end function M,sor

6、t_V,avg_face = averageface()big_array=double(zeros(10304,400);new_big_array=double(zeros(10304,400);avg_face=double(zeros(112,92);i=1; for m=1:40; for k=1:10 FileName = C:UsersAdministratorDesktop编程实验材料att_faces,s,int2str(m),int2str(k),.txt; p = load (FileName); big_array(:,i)=reshape(p,10304,1); %将

7、400张脸转换成10304*400的大矩阵 new_p=reshape(p,112,92); avg_face(:,:)=avg_face(:,:)+new_p(:,:)/400; %求平均脸 i=i+1; endend M=mean(big_array); %对矩阵的每一列求均值 for i=1:400 new_big_array(:,i)=big_array(:,i)-M(i);%去均值后的矩阵end %用SVD求特征向量C0=new_big_array*new_big_array;%求低维度的协方差V0,=eig(C0);k=0;V=new_big_array*V0;%特征向量 for

8、i=1:400 %将特征向量进行归一化10304*400 for j=1:10304 k=k+V(j,i)2; end k=k(0.5); V(1:10304,i)= V(1:10304,i)/k; k=0;end sort_V=double(zeros(10304,400);k=400;for i=1:400 %将特征向量的列按特征值从大到小排列好 sort_V(1:10304,i)=V(1:10304,k); k=k-1; end function tyxs,re_newimage = touyingxishu( sort_V ,avg_face) FileName = C:UsersAd

9、ministratorDesktop编程实验材料att_faces,s,int2str(1),int2str(1),.txt; p = load (FileName); avg_face=double(avg_face); big_array=reshape(p,1,10304); a_big_array=double(reshape(big_array,112,92); big_array=a_big_array(:,:)-avg_face(:,:); big_array=reshape(big_array,1,10304); tyxs=big_array*sort_V; re_newima

10、ge=double(zeros(10304,1); for i=1:400 re_newimage(:,:)= re_newimage(:,:)+tyxs(i)*sort_V(1:10304,i); end re_newimage= reshape(re_newimage,112,92); re_newimage=re_newimage(:,:)+avg_face(:,:); minum=min(min(re_newimage);%对特征向量进行标定 maxum=max(max(re_newimage); d=maxum-minum; re_newimage(:,:)=re_newimage(:,:)-minum; re_newimage(:,:)=re_newimage(:,:)/d; re_newimage(:,:)=re_newimage(:,:)*255; re_newimage=uint8(re_newimage); end 专心-专注-专业