Modeling human false target detection decision behavior in infrared images, using a statistical texture image metric

The ICOM statistical texture image metric incorporates the attributes of global texture matching and of local texture distinctness. The metric is used in this paper to predict human false detection performance (probabilities of false alarms) in both natural and enhanced infrared images, by automatic extraction of the potential false targets in the image. Comparing real experimental data with the metric products reveals very good agreement. Following this result, the metric is used to examine whether the human observer, regarding high and low levels of image clutter, behaves as a constant false alarm rate (CFAR) signal processor, or as a fixed threshold signal processor. It is found that neither one of them is correct. Consequently, a modification to the known CFAR decision behavior model is suggested. The modified model considers the total number of detection decisions (true and false) made by the human observer as the adaptive parameter, instead of the number of only the false detection decisions in the case of the CFAR model. The modified model is tested and confirmed with results obtained both from natural and enhanced images.