Curvature Measures in Visual Information Processing
Open Systems & Information Dynamics
Implicit Calibration of a Remote Gaze Tracker
CVPRW '04 Proceedings of the 2004 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'04) Volume 8 - Volume 08
Robust pupil detection by image difference with positional compensation
VECIMS'09 Proceedings of the 2009 IEEE international conference on Virtual Environments, Human-Computer Interfaces and Measurement Systems
| Hi-index | 0.00 |
Finding the eye(s) in an image is a critical first step in a remote gaze-tracking system with a working volume large enough to encompass head movements occurring during normal user behavior. We briefly review an optical method which exploits the retroreflective properties of the eye, and present a novel method for combining difference images to reject motion artifacts. Best performance is obtained when a curvature operator is used to enhance punctate features, and search is restricted to a neighborhood about the last known location. Optimal setting of the size of this neighborhood is aided by a statistical model of naturally-occurring head movements; we present head-movement statistics mined from a corpus of around 800 hours of video, collected in a team-performance experiment.