A neural network approach to represent raster images by 3-order polynomials

Several approaches have been proposed to transform raster image into vectors. The authors propose a method which uses the characteristics of neural networks and monotonic concave functions to select the optimal windows and control points, then they use the method proposed by T. S. Tsui et. al. (1992) to transform a raster image into vectors. Experiments show that this neural network approach is robust in the presence of noise. >