Artificial neural networks have been shown to perform well in many ima
ge processing applications such as coding, pattern recognition and tex
ture segmentation. In a typical multi-layer model of this class, neuro
ns in each layer are linked by synaptic weights to a receptive field r
egion in the layer below it. The input image itself is linked to the l
owest layer. We propose here a two stage encoder-detector network for
edge detection. The single layer encoder stage, trained in a competiti
ve mode, compresses data from an input receptive field and drives a ba
ck-propagation-trained detector network whose two outputs represent co
mponents of an edge vector. Experimental results show that for the cas
e of step edges in noisy images, the performance of the neural edge de
tector is comparable to that of the Canny detector.