Similarity index beetween different size images

Hi there; I'm looking for a Similarity Index (possibily in [0,1] range) beetween two images of different sizes. I've lookex for corr2, but it wants same size images. Also I've looked for xcorr2, but I don't understand how I could use it to get a similarity index.

Any help?

Answer 1

[Massimo Zanetti]

One idea can be appying a transform that reduce the bigger image into an image of the same size of the smaller one. Then apply any similarity index that works with same sized images.

[Leox91]

Mmm ok, but what kind of transform? 0-padding? I don't know if it will degrade the correlation result.

[Massimo Zanetti]

No! There is a very appropriate function to resize images: http://it.mathworks.com/help/images/ref/imresize.html

By default it uses bicubic interpolation, but you have also other options.

If this answer helped you, please accept it.

[Leox91]

Thank you for your help :)

Answer 2

[Image Analyst]

You can use ssim(), psnr, or immse():

image2 = imresize(image2, size(image1));
ssimval = ssim(image1, image2);

[Leox91]

Thank you for your help, I really found helpful your tutorial :)

[nor azam]

Hi Image Analyst,

i got the same error when i used the given code which the size need to be same.

ref = imread('alif.jpg');

A = rgb2gray(ref);

BW1 = imbinarize(A);

figure

imshow(BW1)

image = imread('alif1.jpg');

B = rgb2gray(image);

BW2 = imbinarize(B);

figure

imshow(BW2)

%to calculate the percentage of similarity between ref and new image

BW2 = imresize(BW2, size(BW1));

ssimval = ssim(BW1, BW2);

This is my code. i already converted the images into binary and i want to find the percentage of similarity between two different size of image. please help me sir.

[Mohammad Farhad Aryan]

Input arguments of ssim() are expected to be one of these types: uint8, uint16, single or double not logical.