Infrared image synthesis of desert backgrounds based on semiempirical thermal models

Abstract

This paper presents an efficient per-pixel technique for synthesizing the IR images of desert backgrounds. The technique is based on semiempirical thermal models using a small database of the statistical features of the model coefficients. It is shown that jointly normal Markov random fields can be used to model these coefficients, if the thermophysical parameters of desert backgrounds are independent normal random Markov fields. A new algorithm is introduced for generating the coefficients with desired covariance matrix and correlation lengths. Simulation and experimental results with the new technique illustrate accuracy comparable to that of the conventional per-pixel approach, while the database reduction is several orders of magnitude.