How to compare the features from an unknown image with the feature vectors I extracted from my database images and display the matched image name?

I stored the features in a mat file:

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

clc;

clear all;

close all;

warning off;

f = processEyeImagesInFolder("F:\8th SEM\Dataset\CASIA");

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

function featureTable = processEyeImagesInFolder(folderPath)

% List all image files in the specified folder

imageFiles = dir(fullfile(folderPath, '*.jpg')); % Update the file extension if needed

% Initialize a cell array to store image names and feature vectors

featureData = cell(numel(imageFiles), 2);

% Iterate through each image file

for i = 1:numel(imageFiles)

% Obtain the full file path

imagePath = fullfile(folderPath, imageFiles(i).name);

% Call the function to get the feature vector

currentFeatureVector = getFeatures(imagePath);

disp(['Image Name: ', imageFiles(i).name]);

disp('Feature Vector:');

disp(currentFeatureVector);

% Store image name and feature vector in the cell array

featureData{i, 1} = imageFiles(i).name;

featureData{i, 2} = currentFeatureVector;

end

% Convert cell array to a table

featureTable = cell2table(featureData, 'VariableNames', {'ImageName', 'FeatureVector'});

% Save the table to a MAT file

save('Feature_Data.mat','featureTable');

end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

my MAT file looks like this:

It is a 9x2 table.

Now how to compare the feature vector obtained from an unknown image, compare it with the feature vectors in the mat file and display the image name with the closest match?

Answer 1

[Balavignesh]

Hello Ridhima,

Based on what I've gathered, you're looking to compare an unknown image against a set of feature vectors that have been saved in your MAT file.

Assuming that the new image file is available in your current working directory, you should start by extracting its feature vectors. This can be done using the same 'getFeatures' function employed for your existing dataset. To measure the similarity between the feature vector of the unknown image and those stored in your dataset, the 'Euclidean distance' metric would be a suitable choice. You'll want to identify the feature vector in your dataset that has the minimum Euclidean distance to the one from the unknown image.

Here's a function crafted to carry out the above process:

% unknownImagePath is the path to your new image.
function closestImageName = findClosestFeatureMatch(unknownImagePath)
    % Load the stored feature vectors from the MAT file
    load('Feature_Data.mat', 'featureTable');
    
    % Extract the feature vector from the unknown image
    unknownFeatureVector = getFeatures(unknownImagePath);
    
    % Initialize variables to store the minimum distance and the index of the closest match
    minDistance = inf;
    closestIndex = -1;
    
    % Iterate over each stored feature vector to calculate distances
    for i = 1:size(featureTable, 1)
        % Extract the current stored feature vector
        storedFeatureVector = featureTable.FeatureVector{i};
        
        % Compute the distance between the unknown feature vector and the current stored vector
        distance = norm(unknownFeatureVector - storedFeatureVector);
        
        % Update the minimum distance and index if the current distance is smaller
        if distance < minDistance
            minDistance = distance;
            closestIndex = i;
        end
    end
    
    % Check if a closest match was found
    if closestIndex == -1
        error('No matches found.');
    else
        % Return the name of the image with the closest feature vector
        closestImageName = featureTable.ImageName{closestIndex};
    end
end

Kindly have a look at the following documentation links to have more information on:

• 'norm' function: https://www.mathworks.com/help/matlab/ref/norm.html
• 'pdist2' function: https://www.mathworks.com/help/stats/pdist2.html

Hope that helps!

Balavignesh