Estimation for the eye movement using the spanning tree of transition probability-based graphs

In this paper, we present an algorithm for estimating the eye movement looking at a picture. Our solution is based on human data measured by a wearable eye tracker device which is able to record the user's eye movement during the record. From the same video streams, measuring the artificial salient points on the image by machine vision algorithms, the clusters of these points are assigned according to human measurement data in a learning procedure, and, as a result, transition probability tables are generated containing information on the behavior of a human subject. Using these data, we constructed an algorithm that estimates the movement of the eye on a picture taken in a similar place and during similar conditions. The algorithm generates graphs based on these probability tables and uses their spanning trees to calculate the position of the points that are likely to be visited by the gaze in human experiments. We compared the results with real human tests where subjects were expected to look at the pictures for a certain time.