Abstract
A method is proposed for estimating a geometrical distortion of an imaging system. This method estimates the geometrical distortion from the change of the image intensity caused by a slight shift of the distortion. This method is based on our experience that when we wave a piece of glass, e.g., a fisheye lens, before an object, we can perceive the optical distortion of the transparent body by the motion of the object pattern. In this paper, we analyze the effect on the image intensity by the shifts of the object and the imaging distortion. An equation concerning the differential coefficients of the distortion, those of image, and image changes by the shifts is derived by Taylor expansion of the image intensity at each point. Then, least-square methods are introduced for the estimation of the differential coefficients of the distortion. Experimental results are presented which demonstrate the proposed method works well for a simulated random dot image. Another experiment using a real scanning electron microscope image which was distorted intentionally by a magnetic disturbance also showed the effectiveness of the method.
