2

I've spent a few days searching a for solution to my problem. I'm having a problem in finding peaks in one large matrix (approx. 1400 x 2700). I need to find the main peak and two smaller peaks - in the other words the three largest peaks in the picture. The matrix is the result of two FFTs of one fits image. The main problem is finding the size of the main peak. I managed to find the maximum of the central peak, but I can't find where is its end due to noise. Does anybody have any ideas or code samples I could look at to find a solution?

Update: And just one more related question: How can I effectively determine threshold ( noise ) in my matrix ? If I could successfully do that, finding peaks will be easy job?

viktor.radovic
  • 508
  • 2
  • 9
  • 20

1 Answers1

0

If you don't care for the peak's size, you might try to apply a moving avarage filter on your matrix first. This way, you will lower your peaks' height on the one hand, but get rid of the noise on the other hand.

After applying the moving avarage filter you might look at the gradient of your matrix, in order to identify the peaks.

theV0ID
  • 4,172
  • 9
  • 35
  • 56
  • Im interested in peaks size.. my main problem is finding position of those 3 peaks. Because sometimes i find main peak, and next two but sometimes i cant due to size of main peak. – viktor.radovic Jun 20 '12 at 08:54
  • Maybe, couldn't you use the described 'blured' matrix (lowering the noise) in order to identify the regions, where the peaks are, and then simply scan for the highest value within this regions? – theV0ID Jun 20 '12 at 09:01
  • I`ll try to use it.. After im done with filtering i identify peaks using gradient? – viktor.radovic Jun 20 '12 at 09:04