Plane-based calibration and auto-calibration of a fish-eye camera

We propose a systematic way for (auto)calibrating a fish-eye lens camera. By taking images of a planar scene with a fish-eye camera, our method automatically estimates the centre-of-distortion (COD), distortion parameters, and other conventional camera intrinsics. We fulfil this by a three-stage algorithm. Each stage accounts for one of the above three calibration tasks. Our main contributions reside in the second stages, in which we design a nine-point minimal solver for the purpose of distortion correction. Our method is applicable to both known planar scene and unknown planar scene. The later case corresponds to an auto-calibration version of the fish-eye camera calibration algorithm.