PCA的应用示例

来源:互联网 发布:易语言手机版远控源码 编辑:程序博客网 时间:2024/04/30 08:03

在 PCA 详细算法介绍 (http://blog.csdn.net/watkinsong/article/details/38536463) 中, 因为篇幅问题 没有给出详细的代码示例, 这里给出代码示例。

通过对人脸图像进行降维深入了解PCA算得使用。

首先看一下数据集, 我们有12张人脸图像, 用10张人脸训练PCA降维矩阵, 剩下的两张可以用作测试。 

   

   

   

需要特别注意: 只能使用训练集样本进行所有的PCA训练过程。

这里所有的代码都采用 octave实现, 跟matlab应该是一致的。


1. 加载图像

%% Initializationclear ; close all; clcfprintf('this code will load 12 images and do PCA for each face.\n');fprintf('10 images are used to train PCA and the other 2 images are used to test PCA.\n');trainset = zeros(10, 32 * 32); % image size is : 32 * 32m = 10; % number of samplesfor i = 1 : mimg = imread(strcat(int2str(i), '.bmp'));img = double(img);trainset(i, :) = img(:);end

2. 特征向量做 normalization


%% before training PCA, do feature normalizationmu = mean(trainset);trainset_norm = bsxfun(@minus, trainset, mu);sigma = std(trainset_norm);trainset_norm = bsxfun(@rdivide, trainset_norm, sigma);

3. 在做特征向量归一化的过程中, 我们为了以后使用归一化参数, 需要保存这些归一化参数。

比如这里可能需要保存mu 和 sigma, 这里我们已 mean face 的方式保存 mu, 因为本示例比较小, 所以没有保存 sigma, 这里保存mu 的目的也仅仅是 为了让大家看一下平均脸的样子。 如果在做项目的过程中, 可能训练PCA是分开进行的, 以后需要进行降维, 那么就需要保存这两个归一化参数。 

%% we could save the mean face mu to take a look the mean faceimwrite(uint8(reshape(mu, 32, 32)), 'mf.bmp');

看一下由10张人脸生成的平均脸:


是不是比较丑?  因为人脸太少了, 再来看看由5000个人脸图像生成的平均脸: 



4. 计算降维矩阵


%% compute reduce matrixX = trainset_norm; % just for convience[m, n] = size(X);U = zeros(n);S = zeros(n);Cov = 1 / m * X' * X;[U, S, V] = svd(Cov);fprintf('compute cov done.\n');

5.  查看特征脸


降维矩阵U中的特征向量, 在关于人脸的降维中,又被称为特征脸,  U 中的每个特征向量相当于找到的降维空间的一个方向。 利用U可以将特征映射到这个空间中。


这里我们把的U中的前几个特征向量保存下来,  看一下特征脸。 U 中的特征向量是按照特征值进行由大到小排序的, 这个排序的顺序也决定了对于降维的影响最大的向量放在最前面。 


       

这里的 eigen face 和人脸的相似性比较高, 因为我们的样本数量比较少, 就10个样本。。。所以会出现这种相似度比较高的情况。


补充: 给出几张用5000张人脸图像训练得到的eigen face, 如下所示:

     


6. 降维

%% dimension reductionk = 100; % reduce to 100 dimensiontest = zeros(2, 32 * 32);for i = 1:2img = imread(strcat(int2str(i + 10), '.bmp'));img = double(img);test(i, :) = img(:);end% test set need to do normalizationtest_norm = bsxfun(@minus, test, mu);test_norm = bsxfun(@rdivide, test_norm, sigma);% reductionUk = U(:, 1:k);Z = test_norm * Uk;fprintf('reduce done.\n');

7. 还原特征(Reconstruction)


%% reconstruction%% for the test set images, we only minus the mean face,% so in the reconstruct process, we need add the mean face backXp = Z * Uk';% show reconstructed facefor i = 1:5face = Xp(i, :) + mu;face = reshape((face), 32, 32);imwrite(uint8(face), strcat('./reconstruct/', int2str(4000 + i), '.bmp'));end%% for the train set reconstruction, we minus the mean face and divide by standard deviation during the train% so in the reconstruction process, we need to multiby standard deviation first, % and then add the mean face backtrainset_re = trainset_norm * Uk; % reductiontrainset_re = trainset_re * Uk'; % reconstructionfor i = 1:5train = trainset_re(i, :);train = train .* sigma;train = train + mu;train = reshape(train, 32, 32);imwrite(uint8(train), strcat('./reconstruct/', int2str(i), 'train.bmp'));end



注: 这里我使用了训练样本为4000张, 因为样本数量太少还原的效果很差。 

看一下特征还原的效果: 左边为原始图像, 右边为还原的图像

对于测试样本还原, 测试样本在降维之前减去了mean face, 所以在还原之后还要加上mean face才是真正的还原的图像。

 

 

 

 


对于训练样本还原, 因为训练样本即减去了mean face 还除以了 standard deviation, 所以在计算得到还原的样本特征后, 还首先要将特征 按元素乘上 standard deviation, 也就是 .* ,  然后再加上mean face才是最后得到的真实的还原的数据。 

看以下几个关于训练样本的还原: 同样左边为原始图像, 右边为还原之后的图像

  

  

  

  



整个工程的全部代码: 


%% Initializationclear ; close all; clcfprintf('this code will load 12 images and do PCA for each face.\n');fprintf('10 images are used to train PCA and the other 2 images are used to test PCA.\n');m = 4000; % number of samplestrainset = zeros(m, 32 * 32); % image size is : 32 * 32for i = 1 : mimg = imread(strcat('./img/', int2str(i), '.bmp'));img = double(img);trainset(i, :) = img(:);end%% before training PCA, do feature normalizationmu = mean(trainset);trainset_norm = bsxfun(@minus, trainset, mu);sigma = std(trainset_norm);trainset_norm = bsxfun(@rdivide, trainset_norm, sigma);%% we could save the mean face mu to take a look the mean faceimwrite(uint8(reshape(mu, 32, 32)), 'meanface.bmp');fprintf('mean face saved. paused\n');pause;%% compute reduce matrixX = trainset_norm; % just for convience[m, n] = size(X);U = zeros(n);S = zeros(n);Cov = 1 / m * X' * X;[U, S, V] = svd(Cov);fprintf('compute cov done.\n');%% save eigen facefor i = 1:10ef = U(:, i)';img = ef;minVal = min(img);img = img - minVal;max_val = max(abs(img));img = img / max_val;img = reshape(img, 32, 32);imwrite(img, strcat('eigenface', int2str(i), '.bmp'));endfprintf('eigen face saved, paused.\n');pause;%% dimension reductionk = 100; % reduce to 100 dimensiontest = zeros(10, 32 * 32);for i = 4001:4010img = imread(strcat('./img/', int2str(i), '.bmp'));img = double(img);test(i - 4000, :) = img(:);end% test set need to do normalizationtest = bsxfun(@minus, test, mu);% reductionUk = U(:, 1:k);Z = test * Uk;fprintf('reduce done.\n');%% reconstruction%% for the test set images, we only minus the mean face,% so in the reconstruct process, we need add the mean face backXp = Z * Uk';% show reconstructed facefor i = 1:5face = Xp(i, :) + mu;face = reshape((face), 32, 32);imwrite(uint8(face), strcat('./reconstruct/', int2str(4000 + i), '.bmp'));end%% for the train set reconstruction, we minus the mean face and divide by standard deviation during the train% so in the reconstruction process, we need to multiby standard deviation first, % and then add the mean face backtrainset_re = trainset_norm * Uk; % reductiontrainset_re = trainset_re * Uk'; % reconstructionfor i = 1:5train = trainset_re(i, :);train = train .* sigma;train = train + mu;train = reshape(train, 32, 32);imwrite(uint8(train), strcat('./reconstruct/', int2str(i), 'train.bmp'));endfprintf('job done.\n');



16 0
原创粉丝点击