auto-binarize a fingerprint

Hi, I am a French student and i'm working on a minutia recognition project. My work is based on this application : http://www.mathworks.com/matlabcentral/fileexchange/16728-fingerprint-application It works with the joined picture but a problem comes with the binarization (if the pixel is a stria -> 0, else -> 1). Indeed, the limit between white and black is manually fixed to 160. I am looking for ideas to make it automatically. Maybe, find the average intensity of the stria... Can someone help me ?

Answer 1

[Jean-Edouard Simonin]

up

Answer 2

[Sean de Wolski]

Perhaps use:

graythresh

to determine a suitable threshold.

[Jean-Edouard Simonin]

I didn't know this fonction. That's a good idea but it's not appropriate to fingerprints. The stria are to close and the result is not usable.

Thank you anyway

[Sean de Wolski]

How are the stria too close? graythresh uses a histogram so location doesn't matter. What class is your data?

[Jean-Edouard Simonin]

It's maybe beacause of the difference of intensity between stria and valley which is not too much strong...

Answer 3

[Jean-Edouard Simonin]

I have an idea but I don't know how to do it. If we draw a histogram of the picture, we can see two Gaussians. To determine the limit between white and black, I have to find the min between these two Gaussians. How could I do that ?

[Sean de Wolski]

It's called the minimum valley method.

Basically find the peaks; it's the bottom of the valley between the two peaks. You could use the histc function and some basic numerical derivatives to find it.