Nonparametric , Model - Based Radial Lens Distortion Correction Using Tilted Camera Assumption ∗

Radial lens distortion prohibits use of simple pinhole camera models in computer vision applications, especially when using wide-angle lenses, which result in barreltype distortion. Usual approach to radial distortion is by the means of polynomial approximation, which introduces distortion-specific parameters into the camera model and requires iterative methods for their calculation. Based on the properties of distorted images, an alternative approach is proposed in this paper. The basic assumption is that distortion occurs due to transformation of the observed differential of radius and is locally dependent of the angle of principal rays. The geometric relations which result from this assumption are complemented with the equations of the perspective radial lens projection function to derive model of radial distortion with single parameter focal length. Experiments were conducted to illustrate the validity and performance of this approach.