I found Circular Hough transform being used to detect eyes/iris in facial images, but that gave a lot of error in identification. There were lot of False positives. Could anyone suggest some modifications or other methods to do so.
There are several methods available in the literature. Bellow I attach some of them:
(a) http://hal.archives-ouvertes.fr/docs/00/37/43/66/PDF/article_ivcnz2006.pdf
(b) http://www.waset.org/journals/ijeee/v3/v3-4-39.pdf
(c) http://ijcsi.org/papers/IJCSI-8-4-1-272-282.pdf
(d) http://journal.info.unlp.edu.ar/journal/journal15/papers/jcst-oct05-3.pdf
(e) http://www.ecse.rpi.edu/~qji/Papers/eye_det_frgc.pdf
Maybe you can try this code http://www.mathworks.com/matlabcentral/fileexchange/13716-face-eye-detection
There are several methods available in the literature. Bellow I attach some of them:
(a) http://hal.archives-ouvertes.fr/docs/00/37/43/66/PDF/article_ivcnz2006.pdf
(b) http://www.waset.org/journals/ijeee/v3/v3-4-39.pdf
(c) http://ijcsi.org/papers/IJCSI-8-4-1-272-282.pdf
(d) http://journal.info.unlp.edu.ar/journal/journal15/papers/jcst-oct05-3.pdf
(e) http://www.ecse.rpi.edu/~qji/Papers/eye_det_frgc.pdf
I should be out of the subject, but if you are allowed to modify the image acquisiton setup, you could use near infrared lighting to simplify the eye detection.
They are some interesting infos here: http://dx.doi.org/10.1016/j.cviu.2004.07.012 (sorry not giving the full text link)
Scale invariant feature transform also detect many points as key features in eye region. Try out vlfeat library. It may help you.
This is a more general resource for computer vision and it might be of help:
http://iris.usc.edu/Vision-Notes/bibliography/contents.html
You may try the eye haarcascades provided with OpenCV library.....but they also may give give false positive detections (but hopefully better than Circular Hough Transform)!!
- Badges
- Science topic