Detección del centro de la pupila usando la transformada de simetría radial

Abstract

Eye-movement analysis is crucial for the functional study of the vestibular system and the diagnosis of neurodegenerative and neuro-psychiatric diseases, among others. Although electro-oculography is the most widely used eye movement recording technique, it requires a complex electronic design to comply with electrical safety standards. On the other hand, video-oculography is more feasible from the point of view of electrical safety, costs and generalization. However, it requires processing methods to accurately detect the pupil center position. In this research work, we implement an algorithm based on the radial symmetry transform to estimate the pupil center position. It is based on gradient calculation in order to detect points of high radial symmetry. It was implemented in MATLAB. The algorithm was evaluated with infrared images from CASIA database. Annotation and pupil center detection processes were carried out using the graphical interface “LABELPUP” developed in this work. The evaluation with 400 images showed competitive results with the state of the art.