A Computerized Method to Diagnose Strabismus Based on a Novel Method for Pupil Segmentation

Abstract

Strabismus is a disease of the human visual system in which the eyes are not properly aligned which results in gazing in different. This paper presents a computer vision driven method for a quantitative decision support in the diagnosis of the disease. An infrared (IR) camera with IR illumination is used to capture the patient eyes. The proposed method estimates the normal vectors of both pupils, and measures the enclosed angle in realtime while the patient gazes at different predefined points. The three main processing steps in the proposed algorithm are: (i) Eye detection using Haar features combined with an Adaboost classifier. (ii) A novel method was developed to detect the ellipse of the pupil in the image frame; this method uses a line integral to detect the pupil ellipse in real time. (iii) Estimation of the pupil normal vector based on the parameters of the segmented pupil ellipse. All three methods were evaluated on real human eyes and on an eye model.