Multiple view geometry in computer visiond
Multiple view geometry in computer visiond
Lens distortion calibration using point correspondences
CVPR '97 Proceedings of the 1997 Conference on Computer Vision and Pattern Recognition (CVPR '97)
Radial Basis Functions
Nonmetric Lens Distortion Calibration: Closed-form Solutions, Robust Estimation and Model Selection
ICCV '03 Proceedings of the Ninth IEEE International Conference on Computer Vision - Volume 2
| Hi-index | 0.00 |
Many of the algorithms in computer vision are based on the assumption of linear imaging process, i.e. the world point, image point and the camera center are collinear. However, this assumption is sometimes violated in the case of real lenses. This phenomenon, known as radial lens distortion, can easily be characterized by straight scene lines imaged as curved lines. The most frequently used remedy for this distortion is to model it using some simple equations. The parameters of these equations are then computed by various methods and reported as a part of calibration parameters of the camera under consideration. This paper, however, proposes a novel approach which does not impose any specific model on radial lens distortion. Instead, it tries to learn the distortion pattern from a sample image and then compensate for it. To this end, the proposed method exploits a special class of functions known as Radial Basis Functions to learn the distortion pattern of straight scene lines. Also, some examples have been provided to illustrate the efficiency of the proposed method.