Image processing algorithms for ocular fundus reflectometry

Ocular fundus reflectometry is a technique aimed at the in-vivo measurement of the reflectance of the tissues of the ocular fundus. Studies have demonstrated a correlation between optical and physiological properties of such tissues in humans and the existence of a control mechanism, called neuro-vascular coupling (NC), which adjusts local blood perfusion to support vision-induced neural activity. We developed an instrument for functional imaging of the neural tissues of the ocular fundus based on reflectance measurements to study the NC. The images acquired with the instrument needed processing to work out reflectance time-courses. The algorithm exploited previously requires long computational time, provides poor discrimination of objects and need manual intervention. We have developed a fully automatic algorithm based on differential multiscale framework for the processing of the images of the ocular fundus with reduced computational time. This algorithm is reasonably efficient to determine relative translational displacement (translation and rotation) between the images and also to remove the geometric distortion. Simulation results performed on the fundus images show that differential multiscale framework based image registration reduces computational times up-to one fourth of the time required by the general purpose algorithm, and provides better alignment precision.