How can I compare one image to multiple images?

For example, I have 200 images in a cell array I=1x200 cells. I want to compare image 1st with the other 199 images to the last image 200th with the other 199 images. I wrote a manual code for ten images, but is there any most straightforward way to do this? Here I used SSIM.

%% Save Images to the Directory

I={};

for kk = 1:NumberOfImages

I{kk} = imread(augimdsFolderLocation.Files{bigidxK(kk)});

end

%% Compare Images

SSIM_I_1=[];

SSIM_I_2=[];

SSIM_I_3=[];

SSIM_I_4=[];

SSIM_I_5=[];

SSIM_I_6=[];

SSIM_I_7=[];

SSIM_I_8=[];

SSIM_I_9=[];

SSIM_I_10=[];

for kk = 1:NumberOfImages

SSIM_I_1(kk,1) = ssim(I{1},I{kk});

SSIM_I_2(kk,1) = ssim(I{2},I{kk});

SSIM_I_3(kk,1) = ssim(I{3},I{kk});

SSIM_I_4(kk,1) = ssim(I{4},I{kk});

SSIM_I_5(kk,1) = ssim(I{5},I{kk});

SSIM_I_6(kk,1) = ssim(I{6},I{kk});

SSIM_I_7(kk,1) = ssim(I{7},I{kk});

SSIM_I_8(kk,1) = ssim(I{8},I{kk});

SSIM_I_9(kk,1) = ssim(I{9},I{kk});

SSIM_I_10(kk,1) = ssim(I{10},I{kk});

end

%% Remove self comparsion

SSIM_I_1(1,1)=0;

SSIM_I_2(2,1)=0;

SSIM_I_3(3,1)=0;

SSIM_I_4(4,1)=0;

SSIM_I_5(5,1)=0;

SSIM_I_6(6,1)=0;

SSIM_I_7(7,1)=0;

SSIM_I_8(8,1)=0;

SSIM_I_9(9,1)=0;

SSIM_I_10(10,1)=0;

%% Compare

[Maximum_Image_1,Index_Image_1] = max(SSIM_I_1);

[Maximum_Image_2,Index_Image_2] = max(SSIM_I_2);

[Maximum_Image_3,Index_Image_3] = max(SSIM_I_3);

[Maximum_Image_4,Index_Image_4] = max(SSIM_I_4);

[Maximum_Image_5,Index_Image_5] = max(SSIM_I_5);

[Maximum_Image_6,Index_Image_6] = max(SSIM_I_6);

[Maximum_Image_7,Index_Image_7] = max(SSIM_I_7);

[Maximum_Image_8,Index_Image_8] = max(SSIM_I_8);

[Maximum_Image_9,Index_Image_9] = max(SSIM_I_9);

[Maximum_Image_10,Index_Image_10] = max(SSIM_I_10);

Answer 1

[yanqi liu]

clc; clear all; close all;
db = fullfile(matlabroot,'toolbox\images\imdata');
files = ls(fullfile(db, '*.jpg'));
NumberOfImages = min(5,size(files,1));
%% Save Images to the Directory
I={};
for kk = 1:NumberOfImages
    I{kk} = imread(fullfile(db, strtrim(files(kk,:))));
end

%% Compare Images
SSIM_A = [];
for kk = 1:NumberOfImages
    for ss = 1:NumberOfImages
        if kk == ss
            continue;
        end
        a = I{kk}; b = I{ss};
        if ndims(a) == 2
            a = cat(3, a, a, a);
        end
        if ndims(b) == 2
            b = cat(3, b, b, b);
        end
        if ~isequal(size(a), size(b))
            b = imresize(b, [size(a,1) size(a,2)], 'bilinear');
        end
        if ~isa(a,'uint8')
            a = im2uint8(a);
        end
        if ~isa(b,'uint8')
            b = im2uint8(b);
        end
        SSIM_A(kk,ss) = ssim(a,b);
    end
end

%% Remove self comparsion

%% Compare
for kk = 1:NumberOfImages
    [Maximum_Image_kk,Index_Image_kk] = max(SSIM_A(kk,:));
    fprintf('\n%d image, the max ssim is %d and the value is %.2f', kk, Index_Image_kk, Maximum_Image_kk);
end

result

1 image, the max ssim is 5 and the value is 0.17
2 image, the max ssim is 4 and the value is 0.18
3 image, the max ssim is 2 and the value is 0.15
4 image, the max ssim is 5 and the value is 0.24
5 image, the max ssim is 4 and the value is 0.24

[KH TOHIDUL ISLAM]

Thank you so much!

Answer 2

[DGM]

Consider the example:

% test images
A = imread('cameraman.tif');
I = {A,imnoise(A,'gaussian'),imnoise(A,'gaussian'),imnoise(A,'gaussian'),imnoise(A,'gaussian')};
NumberOfImages = numel(I);

% build ssim for all images, skipping self-comparison
allssim = nan(NumberOfImages);
for kest = 1:NumberOfImages
    for kref = 1:NumberOfImages
        if kref ~= kest
            allssim(kref,kest) = ssim(I{kest},I{kref});
        end
    end
end
allssim

allssim = 5×5

NaN 0.3399 0.3372 0.3388 0.3395 0.3399 NaN 0.2449 0.2415 0.2476 0.3372 0.2449 NaN 0.2546 0.2440 0.3388 0.2415 0.2546 NaN 0.2489 0.3395 0.2476 0.2440 0.2489 NaN

maxssim = max(allssim,[],1) % maximum non-self similarity for each image

maxssim = 1×5

0.3399 0.3399 0.3372 0.3388 0.3395

Note that representing the self-ssim as nan on the diagonal isn't accurate (the self-ssim would be 1), but using a placeholder facilitates the maximization, and those values are otherwise not used.

Also, if it's desired to find the index of the image with the highest ssim for a given reference, use the two-argument output syntax for max().

[KH TOHIDUL ISLAM]

Thank you so much!

Answer 3

[Image Analyst]

You probably should not read all of the images into a cell array. You may run low on memory. Read in the first image, then have a loop over the rest of the images. You can put all your ssim processing into the loop and just overwrite the current image when you read it in. It's just common sense management of memory. No need for separate loops to read all the images in to a cell array and then another loop to process them.

[KH TOHIDUL ISLAM]

Thank you for your valuable advice. This is helpful.