3-D Model-Based Image Stabilization Using Multiple Visual Cues.
MARYLAND UNIV COLLEGE PARK CENTER FOR AUTOMATION RESEARCH
Pagination or Media Count:
This paper studies the problem of image stabilization, defined here as the process of generating a compensated video sequence where image motion resulting from camera motion has been partially or totally removed. The scheme combines various visual cues such as points and horizon lines, and relies on an Extended Kalman Filter for the estimation of parameters of interest. We study both calibrated and uncalibrated stabilization cases. We address the issues of local versus global stabilization. We consider the problem of the selection of model dynamics for the estimation of warping parameters and illustrate the use of kinetic models for the selective removal of oscillatory motion. Experimental results from video sequences generated from off-road vehicle platforms show good performance of the stabilization schemes.
- Anatomy and Physiology
- Human Factors Engineering and Man Machine Systems