Characterization of clutter in IR images using maximum likelihood adaptive neural system

The use of neural networks to quantify IR image clutter is described. The characterization of image clutter is needed to improve target detection and to enhance the ability to compare performance of different algorithms using diverse images. The neural network presented is the maximum likelihood adaptive neural system (MLANS). MLANS is a parametric neural network that combines optimal statistical techniques with a model-based approach. It is shown that MLANS is better at image clutter characterization than the traditional quadratic classifier because MLANS is not limited to the usual Gaussian distribution assumption of statistical pattern recognition approaches and can adapt to the image clutter distribution