comparison of two histograms using pdist2

hi, I am having two Images I wanted compare these two Images by histograms I have read about pdist that provides 'chisq' but i think the way i am doing is not correct, and what to do to show the result afterwards because this is giving a black image.

i1=imread('blue_4.gif');i1=i1(:,:,1);
[c1,n]=imhist(i1);
c1=c1/size(i1,1)/size(i1,2);

i2=imread('blue_5.jpeg');i2=i2(:,:,1);
[c2,n2]=imhist(i2);
c2=c2/size(i2,1)/size(i2,2);
d=pdist2(c1,c2,'chisq');

please give me a working example thanks

[RANJITHA H V]

pdist2 is showing error

Answer 1

[Anand]

You almost got it right. Just transpose the histograms. So you're last command should be

d = pdist2(c1',c2');

[Muhammad Ali Qadar]

Anand, Taking transpose gives just one single value, otherwise it gives 255*256 Array, I think for Compared histogram there should be an array. What you think waiting for you suggestion

[Anand]

Why do you think there should be an array when you comparing the distance between 2 histograms?

The chi-square distance between two distributions (normalized histograms in this case) is expected to be a single number and is useful in comparing how similar the histograms of the two images are. At least, that's how I used it in my research.

When you do not transpose the histograms, pdist2 interprets it differently and gives back pair-wise distances for the elements in the histograms. That's not what you want.

[Muhammad Ali Qadar]

@Anand, thank you for Clarification, can you tell me what this value exibits and if this value could be used as threshold to for a test image

[Anand]

The smaller the value, the more similar the histograms of the two images are. So, this value can be used as an indicator of how similar the images are. Note that this is a primitive indicator.

You can set a threshold on this value and find the images most similar to the image being compared against.

[Muhammad Ali Qadar]

Quite Logical thanks a lot

[bhagya]

superb

[Sashank]

I am beginner. Please guide me about How can I know how much similar two images are ?

[Devarshi Patel]

How can we do this for RGB histogram?

Answer 2

[Raghunandan]

Just wanted to let you know that pdist2 in MATLAB uses 'Euclidean' distance by default. If you are looking to get 'ChiSquare' distance, please use the code found here.

http://www.cs.columbia.edu/~mmerler/project/code/pdist2.m

[ndoum ekanga steve willy]

But this doesnt give me any output, i cant even see the similarity.

% read two images Im1 = imread('ckt_1.tif'); Im2 = imread('ckt_3.tif');

% convert images to type double (range from from 0 to 1 instead of from 0 to 255) Im1 = im2double(Im1); Im2 = im2double(Im2);

% Calculate the Normalized Histogram of Image 1 and Image 2 hn1 = imhist(Im1)./numel(Im1); hn2 = imhist(Im2)./numel(Im2);

% Calculate the histogram error

f = sum((hn1 - hn2).^2); f; %display the result to console