Comparing an occurance of colors between 2 images

Hello Community,

I have 2 groups of images (I put one example from each group in the attachments). I am trying to compare the colors occurances between image1 and image2. When I find the unique number of colors for both image, I ended up way too much colors (which I dont want). Is there any logical way to limit color amount? I have read that clustering by kmeans and using delta-e distance would help me, but I am not able to figure it out.

Simply then, I would like to find:

Which colors are present in the images and their occurance (pixel count)?

Which colors are present in image1 but not image2? How can i show this in colorcloud()?

Thank you so much in advanced

[DGM]

That really depends what the purpose of finding the colors is. Unless you can assume certain things about the images or unless you classify the colors in the same way for all images, I see no reason to assume that two images containing two identically-colored regions will necessarily have said regions mapped to the same color class when you reduce the palette.

So define what it means for two images to have the "same color"

[DGM]

A naive example:

% some images
A = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1287660/Image_1.jpg');
B = imread('https://www.mathworks.com/matlabcentral/answers/uploaded_files/1287665/Image_2.jpg');

% get histcounts
Ahist = u8rgbhist(A);
Bhist = u8rgbhist(B);

% generate logical maps of color membership
inA = logical(Ahist);
inB = logical(Bhist);
inAonly = inA & ~inB;
inBonly = ~inA & inB;
inboth = inA & inB;
inneither = ~inA & ~inB;

% number of colors in each set
numinAonly = nnz(inAonly)

numinAonly = 57062

numinBonly = nnz(inBonly)

numinBonly = 40866

numinboth = nnz(inboth)

numinboth = 0

numinneither = nnz(inneither)

numinneither = 16679288

function H = u8rgbhist(inpict)
% generate a 3D histogram of all the colors in a uint8 image
    inpict = im2uint8(inpict);
    [rows,cols,ch,~] = size(inpict);
    if ch~=3
        error('U8RGBHIST: expected input to be a RGB image')
    end
    H = zeros(256,256,256);
    for r = 1:rows
        for c = 1:cols
            idx = inpict(r,c,:) + 1;
            H(idx(1),idx(2),idx(3)) =  H(idx(1),idx(2),idx(3)) + 1;
        end
    end
end

There aren't any colors common to the two images.

Answer 1

[prasanth s]

convert the images to HSV colorspace and use the hue component to clustering.

match the colors using the obtained cluster centers. it will give an approximate matching for closely related colors.

check below code

clear
close all

I1 = imread('Image_1.jpg');
L1=rgb2hsv(I1);
I2 = imread('Image_2.jpg');
L2=rgb2hsv(I2);

n_cluster=20;

hue_1=100*L1(:,:,1);
[idx,C_1] = kmeans(hue_1(:),n_cluster);
K1=reshape(idx,size(hue_1));
[counts_1,bins]=histcounts(K1,n_cluster);

hue_2=100*L2(:,:,1);
[idx,C_2] = kmeans(hue_2(:),n_cluster);
K2=reshape(idx,size(hue_2));
[counts_2,bins]=histcounts(K2,n_cluster);

% find common clusters
C_1=round(C_1);
C_2=round(C_2);
[val,pos]=intersect(C_1,C_2);

for i=1:length(val)
    
M1=K1==pos(i);
pos2=find(C_2==val(i),1);
M2=K2==pos2;
J1=I1;
J2=I2;

cluster1 = bsxfun(@times, J1, cast(M1, 'like', J1));
cluster2 = bsxfun(@times, J2, cast(M2, 'like', J2));

figure,imshow(cluster1)
figure,imshow(cluster2)
end