Skip to Main Content
Fitting geometric models to objects of interest in images is one of the most classical problems studied in computer vision field. As a result of its strong representation power and flexibility, conic is one of the geometric primitives widely used in a large number of image analysis applications, in practice. As opposed to most existing conic fitting methods minimizing the fitting error with the use of the second order polynomial representation, in this paper, we propose a new method that formulates the geometric fitting problem as a process of seeking for the optimal mapping to a bivariate normal distribution model. As a result, some critical disadvantages tightly coupled with those methods following the routine polynomial representation can be well overcome. To demonstrate this, a set of carefully designed comparison experiments is conducted to show the superiority of the newly proposed method to a representative method using the polynomial representation. Additionally, the practical effectiveness of the proposed method is further manifested using a set of real image data with a promising accuracy.
Date of Conference: 3-6 Sept. 2009