Assuming that you have a pair of binary images D and T, to find the number of overlapping pixels, you can use:   

D=segmentedCandidate;

T=groundTruth;

n = nnz(D&T)

If you want to compare different segmentation results with the ground truth, some scoring approaches you could use include the Jaccard index, Hamming distance, correlation, Tversky index, and so on.  Below is an example of calculating the DICE metric - borrowing some from the matlab reference (link below):

D_array=[seg_result1, seg_result2, ..., seg_resultN];

dice=zeros(size(D_array))

for i=1:max(size(D_array))

    dice(i) = 2*nnz(D_array(i)&T)/(nnz(D_array(i)) + nnz(T))

end

plot(dice)   % will show DICE similarity for results

This formula can also be used for video if cast into an update formula wheretarget frames are obtained through spatial or temporal derivatives.

https://www.mathworks.com/matlabcentral/answers/68720-dice-similarity-index-between-segmented-image-and-ground-truth

hi

you can use this for your need:

clear;

clc;

%as the simiularity increase , the snr and psnr will increase two

b=imread('c:\3.tif');

c=imread('c:\4.tif');

d=imread('c:\5.tif');

e=imread('c:\6.tif');

f=imread('c:\7.tif');

g=imread('c:\8.tif');

[peaksnr, snr] = psnr(b, c);

[ppp, sss] = psnr(b, f);

[pppp, ssss] = psnr(b, g);

peaksnrmatris=[peaksnr;ppp;pppp];

snrmatris=[snr;sss;ssss];

fprintf(' The Peak-SNR value is ');

psnrnorm=normc(peaksnrmatris)

fprintf(' The SNR value is ');

snrnorm=normc(snrmatris)

%for dissimularity

bb=1-psnrnorm;

cc=1-snrnorm;

https://uk.mathworks.com/help/images/ref/psnr.html

Hello Jeffrey,

Thanks for your information...

is it necessary to convert RGB image to binary?

I have used n = nnz(D&T) on RGB images. It is giving  n=202208 value. Both of images are having size 260*260 with Region of interest (ROI) having RGB color and background is white..

What should I interpret from value n? is it correct?

Nilima,

'nnz' simply returns the number of non-zero elements in a matrix, so to answer the question you pose; 'is it correct?', one would have to say 'yes'.

Ok.. thanks Mark Toolan...Can I apply nnz on RGB image or gary scale image?

Nilima,

'doc nnz' in the Matlab command window will give you the documentation of the function.  As it returns the number of non-zero elements, I can't see a problem with using RGB or grey-scale images.

Regarding your initial question, if the image returned from your (D&T) shows only your ROI (the 'background' being shown as black), you're good-to-go.

Thanks for the information Mark Toolan...

Hello, Nilima: 

This is a good question. Below is the pesudo Matlab codes:

Let us assume the segmented object has pixel label 1 and the background has pixel label 0.

temp=T+D;

Intersection=temp>1;

union=temp>0;

Dice-Measure=intersection/union;

Thanks Zhenzhou Wang..

But my images are RGB. Converting it to binary, is not acceptable..

For color images how to do?

Why not convert to binary? Anyway, what you want to know is the measure-score and the binary image could do the job.