Fast Modeling of Curved Object from Two Images

Abstract

A method of modeling curved object from two images is proposed. 3D curved object is represented by the standard parametric surface. The silhouettes of the images photographed from the frontal and lateral position can be obtained by user interaction. They are inputted as the linear constraints of the quadratic object function to find the smoothest curved surfaces of the object. In order to reduce the amount of user interaction, Gaussian mixture model based image segmentation algorithm is used to obtain the silhouettes. We demonstrate the effectiveness of this method in modeling real curved objects.