Color's histogram and histogram's comparasion.

I've been given a set of images showing different football equipements (t-shirts) with different colors. The idea is, for each image, compute its histogram of RGB colors with N bins, and once computed, compute the similiraty between histograms (using either Chi-Square method or Euclidean distance). I am struggling doing both tasks and I am desperate, so I really hope someone could help me. I'll leave the two functions used down here:

% Histogram computation
function [hist_3d] = Calcular_Histograma_rg(image)
    bins = 9;
    R = image(:,:,1);
    G = image(:,:,2);
    B = image(:,:,3);

    hist_3d = zeros(bins, bins, bins);
    [n m] = size(R);
    for i = 1:n
        for j = 1:m
            r = R(i, j)/ (256/bins) + 1;
            b = B(i, j)/ (256/bins) + 1;
            g = G(i, j)/ (256/bins) + 1;
            if (r > bins)
                r = r - 1;
            end
            if (b > bins)
                b = b - 1;
            end
            if (g > bins)
                g = g - 1;
            end
            hist_3d(r, b, g) = hist_3d(r, b, g) + 1;
        end
    end
    hist_3d = hist_3d ./ (n*m);
    hist_3d = hist_3d ./ (n*m);
end

% Similarity using euclidean distance computation
function [similarity] = Calcular_Similitud_rg(hist_org,hist)    
    [n, m] = size(hist_org);
    dist = 0;
    for i=1:n
        for j=1:m
            aux = (hist_org(i, j) - hist(i, j));
            dist = dist + (aux*aux);
        end
    end
    similarity = sqrt(dist);
end

[Pol]

Maybe it is better to do it in HSV space. Give the appropiate change as you want.

Answer 1

[Image Analyst]

You're going to have to round r, g, and b because they're not integers and can't be used as indexes.

Don't use image as the name of your variable since it's already a built-in function.

Have you thought about what to do if the two images are not the same size? Maybe you want to normalize the histograms.

Answer 2

[LeoAiE]

Maybe you have implemented the histogram computation and similarity calculation using Euclidean distance. there are a few modifications needed. First, in the Calcular_Histograma_rg function, you have duplicated the normalization step, which should only be done once. Second, you need to adapt the Calcular_Similitud_rg function to handle 3D histograms.

% Histogram computation
function [hist_3d] = Calcular_Histograma_rg(image)
    bins = 9;
    R = image(:,:,1);
    G = image(:,:,2);
    B = image(:,:,3);
    hist_3d = zeros(bins, bins, bins);
    [n, m] = size(R);
    for i = 1:n
        for j = 1:m
            r = floor(R(i, j) / (256/bins)) + 1;
            b = floor(B(i, j) / (256/bins)) + 1;
            g = floor(G(i, j) / (256/bins)) + 1;
            hist_3d(r, b, g) = hist_3d(r, b, g) + 1;
        end
    end
    hist_3d = hist_3d ./ (n * m);
end

% Similarity using Euclidean distance computation
function [similarity] = Calcular_Similitud_rg(hist_org, hist)    
    [n, m, p] = size(hist_org);
    dist = 0;
    for i = 1:n
        for j = 1:m
            for k = 1:p
                aux = (hist_org(i, j, k) - hist(i, j, k));
                dist = dist + (aux * aux);
            end
        end
    end
    similarity = sqrt(dist);
end

image1 = imread('image1.jpg');
image2 = imread('image2.jpg');

hist1 = Calcular_Histograma_rg(image1);
hist2 = Calcular_Histograma_rg(image2);

similarity = Calcular_Similitud_rg(hist1, hist2);

[Image Analyst]

Don't use image as the name of your variable since it's already a built-in function.