Abstract

This paper deals with a recently proposed nonparametric approach to camera calibration, which is applicable to any type of sensor design. Currently, no relative quantitative performance data is available for this method. This paper addresses this issue, by providing a comprehensive evaluation with respect to the standard planar calibration technique in the literature. Experiments are conducted on simulated and real data, with the firm conclusion that the generic calibration method has the capability to outperform the standard parametric approach for imaging systems with significant distortion. The results provide important practical information for the vision community at large.