Mosaicing the Retinal Fundus Images: A Robust Registration Technique Based Approach

Abstract

Mosaicing fundus retinal images is fundamental to reveal helpful information of the eyes in order to track the progress of possible diseases. We propose the use of a simple rigid model to globally match vascular trees via a multi-resolution scheme. An elastic matching algorithm is employed to achieve accurate local alignment. We build mosaic maps by merging gray intensities of pixels from different fundus images at the same transformed locations with arithmetic average operation. Experiment results show that successful matching can be achieved with improved accuracy and the mosaic maps seem perfect in terms of visual inspection.