Superresolution of targets on the multi-grade scene: a spectral positional invariance-based approach

We introduce a new technique for resolution of point-type targets on multi-grade background scenes. Among the parametric spectral estimators, the Prony model and Pisarenko harmonic decomposition method are ones most widely used for the characterization of the sinusoidal signal models, but face problems when signals are embedded in colored noise. Nevertheless, the proposed extension of the Prony method permits us to highlight an important fact: under different hypotheses about the number of actual sinusoidal components, the true components remain unaltered in position, while the other spectral lines change. We use this fact to decompose the problem under two paradigms: Prony estimation for point-type targets, and nonparametric maximum entropy estimation applied to reconstruct the image of the background scene. By fusing these two methods, we achieve a substantial improvement in resolutions of point-type targets as well as the background spectral characterization.