Process an image depending of the peaks of its histogram

I'm looking for a matlab function that let me analyse different regions of an image depending of the peaks of its histogram.

I'm working with skin images (that is why I can't upload a picture now to explain better). I'm calculating the mean intensity and standard deviation of each channel of the RGB channel. I take picture of patients that have skin lesions and I process their images along time, to see through the mean of their image how they evolve.

For example, I have the histogram of one of the images,

 and as you can see, I have three peaks of intensity. What I want is to run in this case, three different analysis centring in each peak. Because I found that each peak represented a different region in the image, as well as the pixel near this high values.

So finally I can analise something like this (sorry for the low quality),

At last, as I know that each gaussian represent a region of the image, how could I show after analyzing the image, which part belong to who?

I hope you can understand my doubt!

Answer 1

[Shounak Shastri]

"I'm looking for a matlab function that let me analyse different regions of an image depending of the peaks of its histogram."

The short answer to this question is that you cannot.

Histograms tell you the number of pixels for a particular intensity value and not their location.

For example, Consider these 2 images of size [3, 3].

 A = [3 5 7 ; 4 1 3 ; 4 3 5];
 B = [1 3 4 ; 5 4 3 ; 7 3 5];

If you take the histogram of these two images, you would get the same figure and values and you wont be able to tell which image the histogram belongs to or where are the pixels with intensity value "5" located.

"I take picture of patients that have skin lesions and I process their images along time, to see through the mean of their image how they evolve."

This problem is usually solved by segmenting the lesions from the image and calculating its area. Based on the increase or decrease in the area, you can see if the patient is getting better or not.