Some results on edge enhancement with neural networks

Presents some results on edge enhancement with artificial neural networks. The main motivation of this work is to show, experimentally, that nonlinear filters implemented with neural networks can be superior to the commonly used linear filters. Frequent problems that are found in traditional approaches are edge misplacement, poor handling of corners and blurring (or even suppression) of edges. The implementation described exhibits a reasonable idea of the capacity of neural networks to reduce or avoid some of these drawbacks, both on synthetic and natural images.<<ETX>>