After PCA, the reconstructed image is darker than the original.

Question

예림 2024-5-20

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2120606-after-pca-the-reconstructed-image-is-darker-than-the-original

评论： Image Analyst 2024-5-21

在 MATLAB Online 中打开

original iris image

pca iris image

After PCA of the iris image, I reconstructed the image and it became dark. And when I analyzed the original image and PCA image with SSIM, the result was 0.09333. Other than that, I tried MSE and PSNR, and all the results showed low structural similarity between images. What part of the code should I modify to solve this problem?

clc; clear;
% Importing an iris image
i_image = '홍채 이미지 예시2.jpg';
i = imread(i_image); 
figure(1);
imshow(i); 
title('Original Image');
% Image RGB Channel Extraction
red = double(i(:, :, 1)); 
green = double(i(:, :, 2));  
blue = double(i(:, :, 3));  
% Specifying the number of principal components
num_components = 200;
% Red Channel PCA
[coeff_r, score_r, ~] = pca(red);
reduced_score_r = score_r(:, 1:num_components);
reconstructed_R = reduced_score_r * coeff_r(:, 1:num_components)';
% Green Channel PCA
[coeff_g, score_g, ~] = pca(green);
reduced_score_g = score_g(:, 1:num_components);
reconstructed_G = reduced_score_g * coeff_g(:, 1:num_components)';
% Blue Channel PCA
[coeff_b, score_b, ~] = pca(blue);
reduced_score_b = score_b(:, 1:num_components);
reconstructed_B = reduced_score_b * coeff_b(:, 1:num_components)';
% Reconfiguring the image
reconstructed_red = reshape(reconstructed_R, size(red));
reconstructed_green = reshape(reconstructed_G, size(green)); 
reconstructed_blue = reshape(reconstructed_B, size(blue));
reconstructed_image = cat(3, uint8(reconstructed_red), uint8(reconstructed_green), uint8(reconstructed_blue));
% Show Reconfigured Images
figure(4);
imshow(uint8(reconstructed_image));
title('PCA Image');
% SSIM
ssimval=ssim(i, reconstructed_image);
% MSE
err=immse(i, reconstructed_image);
% PSNR
peaksnr=psnr(i, reconstructed_image);

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Angelo Yeo 2024-5-20

2
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/2120606-after-pca-the-reconstructed-image-is-darker-than-the-original#answer_1460366

在 MATLAB Online 中打开

홍채 이미지 예시2.jpg

The reconstructed image is darker because you did not compensate the mean which is substracted when calculating covariance matrix. See pca for more information.

clc; clear;

% Importing an iris image

i_image = '홍채 이미지 예시2.jpg';

i = imread(i_image);

figure(1);

imshow(i);

title('Original Image');

% Image RGB Channel Extraction

red = double(i(:, :, 1));

green = double(i(:, :, 2));

blue = double(i(:, :, 3));

% Specifying the number of principal components

num_components = 200;

% Red Channel PCA

[coeff_r, score_r, ~,~,~,mu_red] = pca(red);

reduced_score_r = score_r(:, 1:num_components);

reconstructed_R = reduced_score_r * coeff_r(:, 1:num_components)';

% Green Channel PCA

[coeff_g, score_g, ~,~,~,mu_green] = pca(green);

reduced_score_g = score_g(:, 1:num_components);

reconstructed_G = reduced_score_g * coeff_g(:, 1:num_components)';

% Blue Channel PCA

[coeff_b, score_b, ~,~,~,mu_blue] = pca(blue);

reduced_score_b = score_b(:, 1:num_components);

reconstructed_B = reduced_score_b * coeff_b(:, 1:num_components)';

% Reconfiguring the image

reconstructed_red = reshape(reconstructed_R, size(red)) + mu_red;

reconstructed_green = reshape(reconstructed_G, size(green)) + mu_green;

reconstructed_blue = reshape(reconstructed_B, size(blue)) + mu_blue;

reconstructed_image = cat(3, uint8(reconstructed_red), uint8(reconstructed_green), uint8(reconstructed_blue));

% Show Reconfigured Images

figure(4);

imshow(uint8(reconstructed_image));

title('PCA Image');

% SSIM

ssimval=ssim(i, reconstructed_image);

% MSE

err=immse(i, reconstructed_image);

% PSNR

peaksnr=psnr(i, reconstructed_image);

3 个评论
显示 1更早的评论隐藏 1更早的评论

Angelo Yeo 2024-5-20

You can accept the answer if the query is resolved, @예림 :D

Image Analyst 2024-5-21

If this Answer solves your original question, then could you please click the "Accept this answer" link to award the answerer with "reputation points" for their efforts in helping you? They'd appreciate it. Thanks in advance. 🙂 Note: you can only accept one answer (so pick the best one) but you can click the "Vote" icon for as many Answers as you want. Voting for an answer will also award reputation points.

请先登录，再进行评论。