VR 2017

2017 IEEE Virtual Reality (VR), March 18-22, 2017, Los Angeles, CA, USA

Desktop Layout

Poster Session A
Ballroom A/B
Monocular Focus Estimation Method for a Freely-Orienting Eye using Purkinje-Sanson Images
Yuta Itoh, Jason Orlosky, Kiyoshi Kiyokawa, Toshiyuki Amano, and Maki Sugimoto
(Keio University, Japan; Osaka University, Japan; Wakayama University, Japan)
Abstract: We present a method for focal distance estimation of a freely-orienting eye using Purkinje-Sanson (PS) images, which are reflections of light on the inner structures of the eye. Using an infrared camera with a rigidly-fixed LED, our method creates an estimation model based on 3D gaze and the distance between reflections in the PS images that occur on the corneal surface and anterior surface of the eye lens. The distance between these two reflections changes with focus, so we associate that information to the focal distance on a user. Unlike conventional methods that mainly relies on 2D pupil size which is sensitive to scene lighting and the fourth PS image, our method detects the third PS image which is more representative of accommodation. Our feasibility study on a single user with a focal range from 15-45 cm shows that our method achieves mean and median absolute errors of 3.15 and 1.93 cm for a 10-degree viewing angle. The study shows that our method is also tolerant against environment lighting changes.


Time stamp: 2020-03-28T22:08:32+01:00