Automatic Composition of Color Ocular Fundus Images

This paper describes a PC-based system that is able to automatically combine color ocular fundus images captured from different angles by a digital retinal camera. The system matches the locations of retinal vascular nodes (crossing points or branches) in the different images by comparing the correlation values of all nodes in a particular rectangular area with all other nodes in all other rectangular areas. In an experiment involving 28 volunteers, the system successfully generated, in real time, composite ocular fundus montages from the 9 color ocular fundus images taken of each volunteer. It was able to match 246 out of 252 images and automatically generated 23 montages without operator intervention. This system is therefore very promising for the automatic assembly of color ocular fundus montages.