A robust method for extracting and labelling mires in keratographic images

摘要

We are interested in the construction of geometric models for the human cornea. They require the knowledge of a finite number of points, defined on the surface of the cornea, by their position in the 3 D-space. Processing digitized keratographic images gives the elevation of corneal points, with respect to a reference plane, using luminous rings reflected off the cornea. The first step of the process is to correctly extract and label rings points. We propose an efficient method for this aim, based on a geometric transformation of the images, followed by an edge tracking process using a graph search algorithm.