Recovery of Circular Motion Geometry in Spite of Varying Intrinsic Parameters

Previous algorithms for recovering 3D geometry from uncalibrated circular motion image sequences of unknown rotation angles are mostly for constant intrinsic parameters. In this paper, a new and simple method for recovering circular motion geometry in spite of varying intrinsic parameters is proposed. It is shown that the movement of the camera forms two concentric circles on the motion plane. By identifying the concentric conic loci in 3D projective frame, the geometry of circular motion can be recovered. Compared with existing rotation angle recovery methods, the new method is (i) more flexible in that it allows the intrinsic parameters to vary from image to image; (ii) simpler by avoiding the calculation of the intersection points of two conics which is notoriously complicated. Experimental results for real images are provided to show the performance of the proposed method.