Vector quantization based target cueing

Abstract

An imagery based Automatic Target Cueing (ATC) system implies a need for increasingly complex image processing and data handling techniques. Many conventional ATC techniques fall short of handling the volumes of data generated by new sensor systems in a timely and effective manner. This paper describes new Vector Quantization (VQ) based ATC techniques that have resulted in many innovative solutions to basic ATC problems and have advanced the state-of-the-art in crucial areas of image exploitation. In addition, VQ based ATC offers a new look at target recognition, combining elements of image compression and pattern recognition. This paper briefly outlines the theory of Vector Quantization (VQ) based target cueing in the compressed domain. The VQ coding procedure forms vectors from groups of spatial and/or spectral pixel information, and represents those groups of data by reference to a code book. The experiment reported consists of analyses of thirty-six frames of image data. Results are shown in terms of probability of detection versus false alarms per frame.<>