Using MATLAB to find maxima for data with a lot of noise

Question

I have noisy data set with three peaks in MATLAB and want to do some image processing on it. The peaks are about 5-9 pixels wide at the base, in a 50 x 50 array. How do I locate the peaks? MATLAB is very new to me. Here is what I have so far...

For my original image, let's call it array, I tried

J = fspecial('gaussian',[5 5], 1.5);
C = imfilter(array, J)
peaks = imregionalmax(C);

but there is still some noise along the baseline between the peaks so I end up getting a ton of local maxima that are really just noise values. (I tried playing with the size of the filter, but that didn't help.) I also tried

peaks = imextendedmax(C,threshold);

where the threshold was determined visually... which works but is definitely not a good way to do it since it's not that robust obviously.

So, how do I locate these peaks in a robust way?

Answer 1

Quick suggestions:

Try working with a median filter in matlab medfilt2, it is more efficient in removing noise than gauss filter. The Gauss conovlution filter works better with fine noise and preserves the image more.

Then after you extract the peaks they are still not classified, you should classify each peak and decide if it is noise or the peak expected. I suggest you look into the binary image class of functions. Especially look at bwconncomp.