An algorithm of camera self-calibration

This paper presents a method of camera self-calibration based on structural information of scenes containing isosceles trapezoids. First, we establish the homography between a plane in the scene and that of an image and provide the expression of the absolute conic in space under an affine coordinates system. Then using the homography and the constraints of circular points on camera intrinsic parameters, we construct a group of nonlinear equations and determine the camera intrinsic parameters by Levenberg-Marquardt algorithm. The experimental results of both synthetic data and real-world data show that our method possesses a high accuracy.