| Abstrak/Abstract |
Iris center detection is a part of eye tracking research. Instead of using a special eye tracking device, iris center under visible light can be used as an input for a low-cost consumer-grade webcam eye tracker. Detecting iris under visible light, however, is a challenging task. Previous works have developed an iris detection method mostly by implementing Circle Hough Transform. However, the accuracy of Circle Hough Transform decreases when the head is not orthogonally positioned to the camera. To deal with this problem, this paper proposes an improved Circle Hough Transform by implementing morphological image processing in the eye's region of interest to determine the selected iris candidate from the accumulator's threshold of the Circle Hough Transform. Using the proposed method, we could detect iris center during various head poses in more than 83% out of 500 images under 0.25 of normalized error. In this case, we achieved 15.2% higher detection rate compared with iris detection based on Circle Hough Transform. Despite of this improvement, future research needs to be performed to achieve more accurate results during head shifting. |
