Real-Time Optical Edge and Corner Tracking at Subpixel Accuracy
Stefan Brantner5, Thomas Auer5 
and Axel Pinz5
| (5) |
Computer Graphics and Vision, TU Graz Münzgrabenstr. 11, A-8010 Graz, Austria |
Abstract
High position and orientation accuracy is required in augmented reality systems.This paper presents an extension to the well
known XVision library for real-time tracking. We implement a real-time sub-pixel accuracy edge tracking algorithm estimating
edge position and orientation based on “virtual edge” localization in a 5 x 5 window. In addition, we give a simple yet efficient
heuristic for identifying corners. Experiments demonstrate results for sub-pixel edge and corner tracking superior to the
original XVision primitives.
Keywords real-time tracking - augmented reality - subpixel accuracy
This work has been supported by the Austrian Science Fund (FWF) under contract number P-12074-MAT. The “Studierstube” is a
joint project with the Vienna University of Technology. We want to thank Joachim Steinwendner for his input in designing the
sub-pixel tracking algorithm.
References secured to subscribers.