Fast radial symmetry transform to distinguish the pupil

Hi there, i've read the paper about fast radial symmetry transform(Loy and Zelinsky) but don't understand well, i wonder if anybody who uses this method to distinguish the pupil in a picture? Many thanks!

Answer 1

[Image Analyst]

It's not likely that anyone here has read some specific paper. If you still cannot understand what they did, you can try to use your own simpler method, though it may not be as good or robust as their method. For example, if you have the centroid of the pupil, and the radius, you can get the circle circumference coordinates using the FAQ. Then for each of those points on the circumference, use improfile() to get a profile from there to the center. Then determine if it's symmetrical enough using some method, like if the black pupil endpoints, 180 degrees opposite from each other on the pupil, are the same distance away from the center, or whatever you want to determine symmetry.

[Weiqian Jiang]

thanks for your suggestion, I've used the gray-distribution thresholding method to distinguish the pupil boundary, but it's not robust due to unideal lighting conditions, so I need more time adjusting my threshold to cover the most reliable boundary and it's time-costing(about 40 seconds for each frame). I noticed that some paper said using this fast radial symmetry method can get the pupil quickly and robustly(even for automatical use for all frames),so I wanna to have a try on this new method. Thanks for your advice again.