ROBUST PUPIL CENTER ESTIMATION USING INPUT ESTIMATION BASED GAUSSIAN SIGNIFICANCE TEST

摘要

In this paper a low cost pupil center locating system was introduced and target maneuver onset detection algorithm, input estimation based Gaussian significance test, was applied to estimate pupil center. Firstly, eye images were converted into Gaussian significance space of input estimate. In statistical significance space, pupil area was enhanced while other area was depressed. Secondly, threshold selected by input estimation based Gaussian significance test algorithm was used to binarize image in the space of statistical significance. At last pupil center was obtained by calculating mass center of the binarized image. Experiments showed that the proposed method was effective.