Circle has been used as one of the most important features for many computer vision applications, such as object recognition and camera calibration. However, its center point cannot be determined accurately under the projective transformations. This thesis develops a new invariant property of circle, which is invariant under the projective transformations to determine the exact projected position of the circle center using concentric circle patterns.
The new invariant property is applied. Specifically, three algorithms to estimate the projected point of the circle center geometrically are proposed: two algorithms based on the center of ellipses, and the other one using equations of ellipses.
A novel camera calibration algorithm is proposed using the exact feature position extraction using concentric circles. The proposed algorithm estimates the intrinsic camera parameters using to images obtained from the same viewpoint.
The performance of the method is validated using both synthetic and real images. The method shows a comparable performance with respect to similar calibration methods using a planar calibration patterns.