Estimation of gaze for human computer interaction

摘要

Gaze of the human eye may act as a communication medium for limb disabled people to interact with computers. In order to determine gaze direction, the first necessary step in image processing is segmentation of iris from entire eye image. In this paper Daugman's algorithm is implemented that uses Integro Differential Operator (IDO) to differentiate iris boundary from sclera part of eye and to find the centroid, to determine where the person is looking. Also the computation time required for IDO to scan the image is minimized by applying the operator only to the local minima pixels in the image. The algorithm is implemented in MATLAB and tested on static eye images selected from the CASIA database as well as on the eye images captured by an iPhone 4s, 8 MP camera with a resolution of 3264×2448 pixels. This method can be useful in eye tracking to handle the input devices of computer like keyboard and mouse.