An Automatic Approach for Radial Lens Distortion Correction From a Single Image

Abstract

Radial lens distortion correction is a fundamental task in photogrammetry and computer vision disciplines, mainly when metric and quality results are required. Radial lens distortion estimation plays an essential role in several photogrammetric tasks such as: registration of sensors, 3-D reconstruction, and mapping of textures. This paper presents an iterative numerical approach for the automatic estimation and correction of radial lens distortion using only a single image. To this end, several geometric constraints (rectilinear elements and vanishing points) are considered. The reported experimental tests indicate that within certain limits, results from a single image can stand the comparison with those from multiple image bundle adjustment.