How to Enhance this finger print?

Question

I am trying to extract finger print from this image by using green LED backlight but when it convert to binarize image most of finger print is gone.

gray= green

I use this fomular to convert from this color image

to this gray image

I use otsu for binarization. So Which preprocessing I should use before convert to binarization to emphasis the finger print.

you could just upload to imgur.com and link to that image from here, we'll happily include the link for you :) (oh by the way, I don't think the green and grey are exactly the same image) — Marcus Müller, Mar 30 '17 at 08:53
I'm no image processing expert, but I remember Otsu's method being very cluster-oriented. Have you tried other methods, and if so, which? — Marcus Müller, Mar 30 '17 at 08:54
@MarcusMüller Thank you for your respond :)
I know only otsu. Do you have another methods name? I use python+opencv. — Sarin Suriyakoon, Mar 30 '17 at 08:59
These are three different finger print pictures. Could you please take one and post green, grey and binarized? Not doing that makes this more guesswork than it needs to be. — Marcus Müller, Mar 30 '17 at 09:00
Playing a bit around with ImageJ, your image is not uniformly enlightened. Hence, you cannot use a global threshold, such as Otsu's method. You can have a try with local/adaptive thresholding the gray image. See e.g. http://homepages.inf.ed.ac.uk/rbf/HIPR2/adpthrsh.htm or http://dsp.stackexchange.com/questions/2411/what-are-the-most-common-algorithms-for-adaptive-thresholding — Maximilian Matthé, Mar 30 '17 at 11:38
Also, you can have a look at scale-space methods, such as http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=6065496 or http://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/Combined_Slides/13-Scale-Space-Combined.pdf — Maximilian Matthé, Mar 30 '17 at 11:41
yeah, I think you might need to go much more for edge-based approaches. Notice how the background is brighter than the fingerprint on the left of the print, but darker on the right? That's a pretty unfortunate setting, because you can't even say that within a small local patch the fingerprint is the lighter/darker part. If you can change something about the way these photos are made, I'd definitely try to enhance the evenness of background irradiation — Marcus Müller, Mar 30 '17 at 13:02

score 2 · Answer 1 · answered Mar 30 '17 at 18:45

Two steps,

do a simple histogram equalization to make the brightness a little bit more even.
Then use canny edge detector (as suggested by Marcus Muller).

matlab code:

I=imread('Your Image');
G1=rgb2gray(I);
G2=histeq(G1);
E2=edge(G2,'canny'); 
figure imshow(E2,[]);

Result (I hope it is what your are looking for):

How to Enhance this finger print?

1 Answers1