Difficulty finding euclidean distance to match
Show older comments
I have one query image and 9 images in the MATLAB database--I have to find the Euclidean distance between two images and represent them in a matrix. I have done the following:
G=imread('pout.tif');
H = adapthisteq(G,'clipLimit',0.01,'Distribution','rayleigh');
[c1,s1]=wavedec2(H,1,'db1');
disp(c1);
X=c1;
figure,imshow(G);
figure,imshow(H);
fileFolder=fullfile(matlabroot,'toolbox','images','imdemos');
dirOutput=dir(fullfile(fileFolder,'*.tif'));
fileNames={dirOutput.name}
n=numel(fileNames);
for k = 12 : 20(dirOutput)
fileNames1=strcat('fullfile(fileFolder)',fileNames(k))
I = imread(fileNames{k});
J = adapthisteq(I,'clipLimit',0.01,'Distribution','rayleigh');
[c2,s2]=wavedec2(J,1,'db1');
disp(c2);
Y=c2;
E_distance = sqrt(sum((H-J).^2));
figure,imshow(I);
figure,imshow(J);
end
But I get the following error:
??? Error using ==> unknown
Matrix dimensions must agree.
Error in ==> pihu at 20
E_distance = sqrt(sum((H-J).^2));
16 Comments
Walter Roberson
on 24 Mar 2013
What does size(H) show? What does size(J) show? What reason do you have to expect that pout.tif will be exactly the same size image as the other *.tif will be?
diva dave anthony
on 24 Mar 2013
Chandra Shekhar
on 24 Mar 2013
Try to resize both the images H and J before finding Equlidean distance.
diva dave anthony
on 24 Mar 2013
Chandra Shekhar
on 24 Mar 2013
Use imresize command and both image should be equal size.
type help imresize in command window.
diva dave anthony
on 24 Mar 2013
Chandra Shekhar
on 24 Mar 2013
[row col]=size(H); [row1 col1]=size(J);
for example row=480 and col=640 for image H and row1=450 and col1=600 for image J. then you have to resize image J same size as image H.
diva dave anthony
on 24 Mar 2013
Image Analyst
on 24 Mar 2013
Let me give you the formula that is more robust and able to handle color images also:
[rows columns numberOfColorChannels] = size(yourImage);
the formula Chandra gave you will work only for gray scale images. This one will work for grayscale or RGB.
diva dave anthony
on 24 Mar 2013
diva dave anthony
on 24 Mar 2013
Walter Roberson
on 24 Mar 2013
size() gives the actual size, in pixels for the rows and columns.
If you mean "actual size" in some real-world units such as "feet", then you need some external calibration information that cannot be determined from the image alone. But Euclidean distance is not restricted to use with real-world units, so it is unlikely you would need that.
diva dave anthony
on 24 Mar 2013
Edited: Image Analyst
on 24 Mar 2013
Walter Roberson
on 24 Mar 2013
imresize(H,1) is going to resize by a scale of 1, which will leave the size exactly the same. You need to instead resize the images to be the same number of pixels, using the second syntax for imresize()
imresize(H, [rows columns])
Image Analyst
on 13 Apr 2013
Editing away your question is very very much despised and will lead to people not answering your questions in the future.
Randy Souza
on 26 Apr 2013
I have restored the original text of this question.
diva dave anthony: this question has a clear subject and an accepted answer, so it may be valuable to someone else in the future. If you have a good reason why it should be removed from MATLAB Answers, please flag the question, explain why it should be deleted, and an administrator or high-reputation contributor will consider deleting the question. Please do not simply edit your question away.
Accepted Answer
Image Analyst
on 24 Mar 2013
0 votes
What you're calling the Euclidean distance is the RMS difference in intensity between the two images. And that (and PSNR and MSE) is not such a great way to compare images anymore, unless you're looking for a fairly exact match - like you know for a fact that the test image is definitely one of the database images. You'll probably find SSIM is used more often: http://en.wikipedia.org/wiki/SSIM. But I can't run your code because I don't have the Wavelet Toolbox. I've added it to the product list to alert anyone else who might try to run your code.
27 Comments
diva dave anthony
on 25 Mar 2013
Image Analyst
on 25 Mar 2013
What "dis code"?
Walter Roberson
on 25 Mar 2013
Did you fix the imresize() problem ?
diva dave anthony
on 25 Mar 2013
Image Analyst
on 25 Mar 2013
Code to calculate SSIM? No, I don't have code for that. Your code in your original question? Yes, but it errored out because I don't have the Wavelet Toolbox like I said, though perhaps someone else might have been able to run it. And please answer Walter's question and post your new, corrected code.
diva dave anthony
on 25 Mar 2013
Edited: Walter Roberson
on 25 Mar 2013
Walter Roberson
on 25 Mar 2013
No, you are resizing H to the size it already is. That is going to leave H and J as the same size.
Question: are you sure your tif files are grayscale? If they might not be then you are going to need to adjust your size() calls.
diva dave anthony
on 25 Mar 2013
diva dave anthony
on 25 Mar 2013
Image Analyst
on 25 Mar 2013
See Walter's comment under your code above.
diva dave anthony
on 25 Mar 2013
Walter Roberson
on 25 Mar 2013
Hint: you need to resize something to a size of something else.
diva dave anthony
on 25 Mar 2013
Image Analyst
on 25 Mar 2013
Edited: Image Analyst
on 25 Mar 2013
Use this:
[rows cols]=size(H);
[rows1 cols1]=size(J);
H = imresize(H, [rows1 columns1]); % Resize H to match J.
% No need for intermediate varaibles C and D anymore.
E_distance = sqrt(sum((H-J).^2)); % H and J must be double, not uint8.
diva dave anthony
on 25 Mar 2013
Edited: Image Analyst
on 25 Mar 2013
diva dave anthony
on 25 Mar 2013
Image Analyst
on 25 Mar 2013
I don't see why C and D are even needed. As far as I can tell, they're not, so my code didn't use them.
diva dave anthony
on 25 Mar 2013
diva dave anthony
on 25 Mar 2013
diva dave anthony
on 25 Mar 2013
Walter Roberson
on 25 Mar 2013
E_distance = sqrt(sum((H(:)-J(:)).^2));
diva dave anthony
on 25 Mar 2013
Walter Roberson
on 25 Mar 2013
On the line after the above assignment (note it is not the same as what you had), add the line
fprintf('distance to file "%s" was %g\n', fileNames{k}, E_distance );
Image Analyst
on 25 Mar 2013
What do you mean e.distance??? You don't have a structure "e" with a member (field) called distance. All you have is a variable called E_distance, which is the Euclidean distance, or the RMS difference. So I'm not sure what you're asking. You say "where i will get the result?" Well, it's right there in that variable.
Walter Roberson
on 25 Mar 2013
The calculation is in a loop and the value was not being printed out, so the value would have been disappearing.
diva dave anthony
on 25 Mar 2013
diva dave anthony
on 25 Mar 2013
More Answers (0)
Categories
Find more on Image Sequences and Batch Processing in Help Center and File Exchange
on 24 Mar 2013
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!
