how to set threshold value for comparing images in matlab?

how do I compare one image for example Z with multiple images in a folder and find out the one which matches closely to the image Z.

Answer 1

[Image Analyst]

You develop an algorithm. That's about all I can say, given the (lack of) information you've provided. Please read this and provide more info.

[AFFY]

this is the code I have but I need to make changes in this one. this works only when two images are exactly the same. But I need to modify this to make it acceptable when two images are nearly the same. like suppose if I get some images scanned, not necessarily the scanned results would be the same, so it should work there. clear all; close all; clc;

path='d:\pic'; list=dir([path, '*.jpg']); img=imread('d:\pic'); % image to compare with X1=rgb2ind(img,16); for x=1:length(list) images{x}=imread([path, list(x).name]); X2=rgb2ind(images{x},16); c=corr2(X1,X2); figure(x); subplot(1,2,1),imshow(images{x}), xlabel(['Image ',num2str(x)]); subplot(1,2,2),imshow(img); if c==1 xlabel('The images are same.'); else xlabel('The images are not same'); end end

[Image Analyst]

You can use psnr(), or ssim(). If you don't have a modern version of the Image Processing Toolbox, you can see the attached psnr code. You can also computer the RMS, MSE, or Median Absolute Difference. There are lots of ways to compare images so try them all - they're simple - and pick the one you like best.

[AFFY]

thanks a lot. Is there any way of interfacing a scanner to MATLAB?? So that the images are directly read from the scanner without human intervention??

[Image Analyst]

Not that I'm aware of. You might be able to use a twain interface or java robot but I can't help you there.

Answer 2

[Pranav Shukla]

close all; 
clc;
path='C:\Users\Vinayak Shukla\Documents\MATLAB\MATLAB PROJECTS\MASTER THESIS FOLDER 2023\happy15_encrpted4.bmp';
list=dir([path, '*.jpg']);
img=imread('C:\Users\Vinayak Shukla\Documents\MATLAB\MATLAB PROJECTS\MASTER THESIS FOLDER 2023\happy15_encrpted4.bmp');

%image to compare with X1=rgb2ind(img,16);
forx=1:length(list); 
images{x}=imread([path, list(x).name]);
X2=rgb2ind(images{x},16); 
c=corr2(X1,X2);

figure(x); 
subplot(1,2,1),imshow(images{x}), xlabel(['Image ',num2str(x)]); 
subplot(1,2,2),imshow(img);
if c==1 
    xlabel('The images are same.');  
else
    xlabel('The images are not same');
end 

This is the error i am getting

Unrecognized function or variable 'x'.

Error in pranav17 (line 8)

images{x}=imread([path, list(x).name]);

[Image Analyst]

@Pranav Shukla That's right. You have not defined x before you passed it in to figure because you have this

forx=1:length(list);

instead of this

for x = 1 : numel(list)

There is no "forx" command or function.

Why are you trying to pass something to figure? Just simply do this:

figure;

And there is no need to do correlation, which takes time, when you can simply do

theAreEqual = isequal(X1, X2);

[Pranav Shukla]

images{x}=imread([path, list(x).name]);

in this line a warning is coming and it says variable appears to change size on every loop iteration(within a script. consider preallocating for speed.

how do i preallocate for speed this above line?

[Image Analyst]

Before the loop, do this to aviod the warning:

images = cell(numel(list), 1);

[DGM]

Here. This is cleaned up.

I don't know why you were converting things to indexed color and then trying to compare the index arrays. The mapping is not continuous, so there would be no point in finding the correlation coefficients. They'd either be equal or not; coefficients less than 1 would be meaningless as a measure of image similarity. I simply omitted that.

If you need the images for later, preallocate by creating a cell array. If you don't need them outside the loop, don't bother storing them.

% read from a set of IPT test images
basepath = fullfile(matlabroot,'toolbox/images/imdata');
filenameexpr = 'AT3*.tif';
matchingfiles = dir(fullfile(basepath,filenameexpr));

% reference image
% doing rgb2ind() on images and then comparing them does not
% yield a comparison of the actual image values.
% i'm just going to grab the first image in the sequence
Xref = imread(fullfile(matchingfiles(1).folder,matchingfiles(1).name));

numimages = numel(matchingfiles);
imagestack = cell(numimages,1); % preallocate
for k = 1:numimages
    % read a file to test
    thispath = fullfile(matchingfiles(k).folder,matchingfiles(k).name);
    Xtest = imread(thispath);
    
    % store it if necessary
    % if you don't need to keep them for later, don't bother
    imagestack{k} = Xtest;
    
    % test it somehow (corr2, immse, isequal)
    c = immse(Xref,Xtest); % mean-square error
    
    % generate a pile of figures
    figure(k); 
    subplot(1,2,1),imshow(Xtest), xlabel(['Image ',num2str(k)]); 
    subplot(1,2,2),imshow(Xref);
    if c==0 % if error is zero
        xlabel('The images are same.');  
    else
        xlabel('The images are not same');
    end 
end

[Pranav Shukla]

basepath = fullfile(matlabroot,'toolbox/images/imdata');
filenameexpr = '(C:\Users\Vinayak Shukla\Documents\MATLAB\MATLAB PROJECTS\MASTER THESIS FOLDER 2023\happy15_encrpted4.bmp)';
matchingfiles = dir(fullfile(basepath,filenameexpr));
Xref = imread(fullfile(matchingfiles().folder,matchingfiles().name));
numimages = numel(matchingfiles);
imagestack = cell(numimages,1); % preallocate
for k = 1:numimages
    thispath = fullfile(matchingfiles(k).folder,matchingfiles(k).name);
    Xtest = imread(thispath);
    
    % store it if necessary
    % if you don't need to keep them for later, don't bother
    imagestack{k} = Xtest;
    
    % test it somehow (corr2, immse, isequal)
    c = immse(Xref,Xtest); % mean-square error
    
    % generate a pile of figures
    figure(k); 
    subplot(1,2,1),imshow(Xtest), xlabel(['Image ',num2str(k)]); 
    subplot(1,2,2),imshow(Xref);
    if c==0 % if error is zero
        xlabel('The images are same.');  
    else
        xlabel('The images are not same');
    end 
end

>> pranav17

Error using fullfile

Not enough input arguments.

Error in pranav17 (line 26)

Xref = imread(fullfile(matchingfiles().folder,matchingfiles().name));

could some please rewrite the whole code again in a simplified manner, and post it here.

the code should be how to compare the threshold value of two images?

please help me out.

[Image Analyst]

You have to give it an index of course. Is the reference your first image? If so,

Xref = imread(fullfile(matchingfiles(1).folder,matchingfiles(1).name));