Calibration of HyperOmni Vision based on Conic Curve

In this paper, we propose a calibration method for catadioptric camera sy stems consisting of a rotation symmetry mirror, like H yperO mni V ision, and an affi ne camera. T he proposed method is based on conventional camera calibration and mirror postu re estimation. M any methods for camera calibration have previou sly been proposed. In the last decade, methods for catadioptric camera calibration have also been proposed. T he main problem with catadioptric camera calibration is mirror postu re estimation becau se the degree of freedom of a mirror postu re is limited and the accu racy of the estimated parameters is inadequ ate, owing to nonlinear optimization. O u r method can estimate the six degrees of freedom of mirror postu re and can be free from the volatility of nonlinear optimization. O u r method u ses a conic cu rve in an image, the borderline between mirror and non-mirror and estimate u p to for mirror postu re. T his method is an application of ex trinsic parameter calibration based on conic fi tting. T he mirror postu re estimated analy tically is not u niqu e, so we propose a selection method for fi nding the best one. B ecau se of the conic-base analy tical method, ou r method can avoid the initial valu e problem arising from nonlinear optimization. W e condu ct experiments on synthesized images and real images to test the performance of ou r method, and discu ss its accu racy .