ABSTRACT
This chapter aims to review on the registration and fusion of human retinal images. While registration is an active topic, fusion of ophthalmic images is a relatively new application domain. Images are acquired from 10 sec to 15 min after the injection while the dye circulates through the retinal arteries, capillaries, veins, and is eliminated from the vasculature, while staining in the optic disc and lesions. Retinal image distortions come from different sources: change in patient sitting position, change in chin cup position, head tilting and ocular torsion, distance change between the eye and the camera, three-dimensional retinal surface, and inherent aberrations of the eye and camera optical systems. Registration methods using similarity metrics like the cross-correlation metric and the correlation coefficient metric have been tested for unimodal ophthalmic image registration. Image registration is a necessary step in producing mosaics that display a large area of the retina, combining segmented features on one image, or performing pixel-level image fusion.
