Signal Source Separation of the EGRET Gamma-Ray Sky Map.

We present a decomposition of the EGRET gamma-ray sky map into signal components from different model classes. Several models are used to generate spatial flux distributions for various non-thermal gamma-ray emission processes. Model classes are then derived based on the characteristics of their wavelet decompositions, while maximizing the separability of the various physical components. The model class information is then used to reconstruct a good statistical representation of the EGRET data using a linear combination of "typical" members from each of the classes.