Finding Parametric Curves in an Image

Abstract

We present a reliable and efficient method for extracting simple geometric structures, i.e., straight lines, parabolas, and ellipses, from edge images. The reliability of the recovery procedure which builds the parametric models is ensured by an iterative procedure through simultaneous data classification and parameter estimation. The overall relative insensitivity to noise and minor changes in input data is achieved by considering many competitive solutions and selecting those that produce the simplest description, i.e., the one that accounts for the largest number of data points with the smallest number of parameters while keeping the deviations between data points and models low. The presented method is efficient for two reasons: firstly, it is designed as a search which utilizes intermediate results as a guidance toward the final result, and secondly, it combines model recovery and model selection in a computationally efficient procedure.