A RTIFICIAL neural network (NN) architectures have been recognized for a number of years as a powerful technology for solving real-world image processing problems. The primary purpose of this special issue is to demonstrate some recent success in solving image processing problems and hopefully to motivate other image processing researchers to utilize this technology to solve their real-world problems. Finally, it is our hope that this special issue will increase the awareness of image processing researchers to the impact of the neural network-based algorithms. From the response to the initial call for papers, ten manuscripts have been selected for inclusion in this special issue. Eight papers have been offered as full papers and two as correspondence items. These papers covered the following major topics: 1) neural network-based algorithms for character recognition; 2) automatic target recognition using artificial neural networks; 3) object identification, classification and segmentation; 4) image prediction and compression. The first paper by Garris et al. provides an overview of the NN-based approaches to optical character recognition (OCR). In this paper the authors present results from the evaluation of several NN-based OCR systems. They also provide an end-toend OCR recognition system based on an enhanced multilayer perceptron (MLP) classifier. The next three papers deal with the topic of the automatic target recognition (ATR). The paper by Wang et al. proposes a new ATR classifier based on an NN architecture called the modular neural network (MNN) classifier. The MNN classifier consists of several independent neural networks trained on local features extracted from specific portions of the image. The final classification is achieved by combining the decision produced by each individual neural network by a method known asstacked generalization . This NN-based classifier is tested on a large set of real forward-looking infrared imagery. Young et al. present a method for detecting and classifying a target from a multiresolution foveal image. In this algorithm target identification decisions are based on minimizing an energy function which is implemented by a novel multilayer Hopfield neural network. This energy function supports connections between nodes at the same level as well as interconnections between two sets of nodes at two different