A model-based neural network for edge characterization

In this paper, we investigate the feasibility of characterizing significant image features using model-based neural network with modular architecture. Instead of employing traditional mathematical models for characterization we ask human users to select what they regard as significant features on an image, and then incorporate these selected features directly as training e xamples for the network. As a first step, we consider the problem of the ch aracterization of edges, which are usually regarded as significant image fe atures by humans. Unlike conventional edge detection schemes where decision thresholds have to be specified, the current NN-based edge characterizatio n scheme implicitly represents these decision parameters in the form of net work weights which are updated during the training process, and which thus allow automatic generation of the final binary edge map without further par ameter adjustments. Experiments have confirmed that the resulting network i s capable of generalizing this previously acquired knowledge to identify im portant edges in images nor included in the training set. In particular, on e of the important attributes characterizing the current approach is its ro bustness against noise contaminations: the network can be directly applied to noisy images without any re-training and alteration of architecture, as opposed to conventional edge detection algorithms where re-adjustment of th e threshold parameters are usually required. (C) 2000 Pattern Recognition S ociety. Published by Elsevier Science Ltd. All rights reserved.