Using Singular value decomposition for feature extraction from images

8 次查看（过去 30 天）
Ron Herman
Ron Herman 2020-4-17
回答： Yee Mon 2024-3-13
Suppose I have a single subject who has 10 images under different conditions. I need to apply singular value decomposition on all 10 images for feature extraction. Then storing the common extracted feature from 10 images.
The purpose is to save space. Rather than saving all 10 images I would like to save common extracted in one place and use the common extracted feature for future face recognition of the same subject (person).
The 10 images have been attached.
  2 个评论
Christine Tobler
Christine Tobler 2020-4-17
Do you want to compute the singular value decomposition of each of these images and store those? Or compute a combined decomposition of all 10 images?
Ron Herman
Ron Herman 2020-4-17
编辑：Ron Herman 2020-4-17
SVD on each image for feature extraction.
I also want to know feature extraction using SVD

请先登录，再进行评论。

请先登录，再回答此问题。

回答（2 个）

Christine Tobler
Christine Tobler 2020-4-17
Image compression using SVD is a pretty common example (although not the most efficient way to compress an image), here are some relevant links:
https://www.youtube.com/watch?v=mSYgi0KCLv8
https://www.mathworks.com/matlabcentral/fileexchange/69916-image-compression-using-svd
https://stackoverflow.com/questions/13614886/using-svd-to-compress-an-image-in-matlab
For feature extraction, I don't know of a "standard" algorithm to do this using SVD. I found some papers searching for this topic, maybe one of these could be helpful:
https://scholar.google.com/scholar?q=feature+extraction+using+svd
Yee Mon
Yee Mon 2024-3-13
% Load the image
image1 = imread('yme.jpg');
% Convert the image to grayscale if it's RGB
if size(image, 3) == 3
image = rgb2gray(image);
end
% Perform Singular Value Decomposition (SVD) on the image matrix
[U, S, V] = svd(double(image));
% Choose the number of singular values to keep (feature extraction)
num_features = 50;
U = U(:, 1:num_features);
S = S(1:num_features, 1:num_features);
V = V(:, 1:num_features);
% Reconstruct the image using the selected features
reconstructed_image = unit8(U * S * V');
% Display the original and reconstructed images by SVD features
subplot(1, 2, 1);
imshow(image1);
title('Original Image');
subplot(1, 2, 2);
imshow(reconstructed_image);
title('Reconstructed Feature Image with SVD');

类别

在 Help Center 和 File Exchange 中查找有关 Eigenvalues 的更多信息

标签

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by