Non-Gaussian stochastic model for analysis of automatic detection/recognition
We demonstrate that a very simple stochastic model based on nonlinear transformation of Gaussian random fields can be successfully used to model homogeneous non-Gaussian natural backgrounds observed in a wide range of airborne and spacebome sensor imagery. We use this model to simulate backgrounds ranging from IR forest terrain to SAR woodland and SAR sea surface imagery. The model reproduces the histogram, second-order correlations, and third-order correlations measured in the real imagery. We discuss applications in the design and analysis of algorithms for automatic detection and recognition of objects embedded in natural imagery.
KeywordsSynthetic Aperture Radar Synthetic Aperture Radar Image Gaussian Random Field Simple Stochastic Model Small Target Detection
- 1.Bertilone, R.S. Caprari, S. Angeli and G.N. Newsam, Applied Optics 36 (1997) 9167Google Scholar
- 5.P.B. Chapple and D.C. Bertilone, “Stochastic simulation of infrared non-Gaussian natural terrain imagery,” to appear in Optics Communications (1998)Google Scholar
- 6.G.N.Newsam and M.Wegener, “Generating non-Gaussian random fields for sea surface simulations,” Proceedings 1994 ICASSP, April 19–22, Adelaide, South AustraliaGoogle Scholar