A neural network energy minimization approach to approximation of 2-dimensional shapes
We introduce a novel method for approximating boundaries of two-dimensional shapes using a neural network energy minimization approach. Design of the energy function allows control over the behavior of the system, which ultimately determines whether each point along the perimeter is or is not a vertex. Shapes are approximated with straight segments, circular arcs, and spiral arcs. The procedure is independent of object rotation. The method is evaluated on various shapes.
Unable to display preview. Download preview PDF.
- 1.Freeman, H.: Shape Description Via the Use of Critical Points. Proc. of IEEE Comp. Soc. Conf. on Pattern Recognition and Image Proc. PRIP-77, (1977) 168–174Google Scholar
- 2.Geman, S., Geman, D.: Stochastic Relaxation, Gibbs Distributions, and Bayesian Restoration of Images. IEEE Transactions on PAMI 6 (1984) 721–741Google Scholar
- 3.Hopfield, J.: Proc. Nat. Academy of Sciences USA, 81 (1984) 3088–3092Google Scholar
- 4.Hopfield, J., Tank, D.,: Biological Cybernetics 52, (1985) 141–152Google Scholar
- 5.Koch, C., Marroquin, J., Yuille, A.: Analog neuronal networks in early vision. Proceedings of the National Academy of Sciences, USA, 83 (June 1986), 4263–4267Google Scholar
- 6.Ramer, U.: An Iterative Procedure for the Polygonal Approximation of Plane Curves. Computer Graphics and Image Processing 1 no. 3 (1972) 244–256Google Scholar