Non Parametric Distortion Correction in Endoscopic Medical Images

Endoscopes consisting of a probe mounted with a camera head, are frequently used in medicine for inspection and visualization of human body cavities (knee and shoulder articulations, bronchi, nose, brain, etc). However the images often suffer from strong lens distortions. Estimation and correction of image distortion is important not only to improve the practitioner´s perception of the inspected area but also to develop systems for 3D navigation and computer assisted surgery. In this paper we compare various conventional calibration methods against the new parameter free method proposed by Hartley and Kang. We believe the non-parametric method is more suitable for endoscopic imaging. We present numerical analysis of the goodness of fit of other conventional approaches as well as calibration results on real images. It should be noted that our results are directly applicable to all vision applications using wide-angle lenses.