Coarse-to-fine strategy for ellipse center location

In the practical application of machine vision, highly precise ellipse center location is the key to measurement accuracy of the system. In this paper, we propose a robust and efficient method of ellipse center location in various noise effects. The proposed method exploits a coarse-to-fine strategy, the tangents near a contour vote in a Hough voting space, followed by mean shift method for the candidate ellipse center, and the pixel-level ellipse center is gotten by Hessian determinant. Finally interpolation method is used to obtain subpixel center coordinates for ellipse feature. Experimental results reveal our approach has a good location performance and the location error is less than 0.12 pixels.