Usually the fixed and moving images will not have black triangles because they are your original source images. Once your moving image has been registered (aligned, rotated, translates) it they then have black triangles shifted in.
Your algorithm should be smart enough to find the irises even with the black triangles. There are many iris image processing algorithms here:
- 22.5.3.1 Periocular Biometrics
- 22.5.3.2 Iris Based Biometrics, Biometrics Engineering Research Center, Dongguk University
- 22.5.3.3 Iris Recognition Systems, Infrared, Near-Infrared, NIR Based
- 22.5.3.4 Iris Recognition Systems, Spoofing, Liveness, Presentation Attack, Contact Lenses
- 22.5.3.5 Iris Recognition Systems, Wavelet Based
- 22.5.3.6 Iris Detection, Segmentation and Localization Systems
- 22.5.3.7 Iris Recognition Systems, Systems, Evaluation, Comparison, Surveys
However if some some reason you think you need to get rid of them there are two ways (again, probably neither is needed);
- Find where there are pure black pixel that are in the corner and then replace them with some other gray level, like the mean of the image or something, or
- Find where the runs of black along the edge of the image end and this will give you coordinates for the inside cropping box. Of course once you crop it out it will be a different size than your source images, but the overall size of the image may not matter when you go to find the iris.
Contrast adjustment may of may not be needed. You may need to do it if, once you've found the circular ROI, to match it to one of the reference images in the database you are going to compare it against. But I would definitely NOT use histogram equalization using some function like histeq. That would be a mistake. What you'd want to use is a function like imhistmatch, or even imadjust instead. That way your test image could have the same intensity range as your reference image. Once that is done you can inspect the pattern within the circular ROI to see which individual's pattern in your database it is closest to.
Hopefully you have some sort of controlled apparatus for capturing the photo. If you don't, like if the illumination varies, then the person's pupil could be a different diameter than what it is in their reference photo in the database and that would make it harder to compare patterns in the annular region.
