Target detection in clutter depends sensitively on the spatial structure of the latter. In particular, it is the ratio of the target size to the clutter inhomogeneity scale which is of crucial importance. Indeed, looking for the leopard in the background of leopard skin is a difficult task. Hence quantifying thermal clutter is essential to the development of successful detection algorithms and signature analysis. This paper describes an attempt at clutter characterization along with several applications using calibrated thermal imagery collected by the Keweenaw Research Center. The key idea is to combine spatial and intensity statistics of the clutter into one number in order to characterize intensity variations over the length scale imposed by the target. Furthermore, when properly normalized, this parameter appears independent of temporal meteorological variation, thereby constituting a background scene invariant. This measure has a basis in analysis of variance and is related to digital signal processing fundamentals. Statistical analysis of thermal images is presented with promising results.