Autocorrelation and shift vectors

Question

Basically I'll be using Autocorrelation method to try to find cloned regions within an image.

This is what i did in MATLAB

i = imread ('D:\image.jpg') I = rgb2gray(i); imshow(I); f = fspecial('LOG'); h = imfilter(I,f); x = xcorr2(double(h), double(h)); imagesc(x); figure(gcf)

basically loading up an image, changing it to greyscale, applying a high pass filter on it, and doing autocorrelation.

Original Image http:// i38.tinypic.com/2u7ojdz.jpg

After autocorrelation http:// i38.tinypic.com/2z3nh9w.jpg

Alright here comes the question. I would like to find the shift vector between the biggest dot in the middle and the small white dot at the top. I had an idea in mind now i got a feeling it doesn't work out anymore.

Any tips/ideas on how they would be achieved (in calculating the shift vector)?

I'm kinda new to this whole image processing / matlab thing, so any help will be appreciated

Answer 1

e.g. convert to HSB format use max to find the brightest dot in each row, then do peak detection to find the local extrema. This will give you the coordinates of each of two spots of interest in your image. From those coordinates finding the delta function is easy.