how to detect the hair from the skin images.

QUESTION 1.I want to detect the hairs from the images. I used the guassian kernal and the first order derivative of guassian. It is giving good results on some images but in some images, it is nor working at all. As you can see in the image, for first two images it is giving some results but in the third image, it is also detecting the region of interest. How to solve this problem. QUESTION 2. another question is in the binary image the hair detected are broken and there are some unnecessary details. how to remove the unnecessary details and connect the broken hairs. (I have used the morphological operations but it is not giving me the good results).

I will be really thankful if anyone can help me.

Answer 1

[Image Analyst]

Have you searched for "dull razor", which some people use to remove hairs?

[Nazia Hameed]

Yes, I have implemented the dull razor algorithm. Also, I download the windows version from http://www.dermweb.com/dull_razor/ but it is not working on all images. I tried on several but it just worked for very few images.My dataset is really complex(ISBI dataset). The algorithm I am currently implementing is giving good results so far. But I am stuck with the above-mentioned problems.

[Image Analyst]

Well you've done more than I have with it then. If that doesn't work, all I can suggest it try things like adapthisteq() and thresholding, or a bottom hat filter with imbothat(). Or you might try a Frangi or Hessian filter - see the File Exchange for those filters.

[Nazia Hameed]

Ok thankyou so much Image Analyst.I Will try them and let you know. Really greatful. Stay blessed

[Nazia Hameed]

I have somehow managed to detect the hair (although thin are not detected properly). I have posted another question regarding another kind of noise can you please look and guide me. Link to the question is below.

https://uk.mathworks.com/matlabcentral/answers/359339-how-can-i-remove-the-circles-different-colours-from-the-dermoscopic-image