Nonlinear Enhancement of Visual Target Detection

The establishment of an analytic framework for image processing algorithmic development utilizing stochastic resonance is a prime objective of this proposal. In order to achieve this goal, it is imperative to assess the relationship of the non-Gaussian nature of the noise processes, the non-linear aspects of the signal processing or suboptimal detection device, as well as the characteristics of the sub-threshold signals. An important consideration here is that the non-Gaussian noise probability density function (PDF) may actually provide a potential for improved detection performance provided that an appropriate detection strategy is employed. Specific goals include (a) an assessment of algorithms capable of estimating the PDF's associated with imagery for which such distributions are unknown, (b) to assess visual image fusion considerations for human perception, (c) to assess image quality measures, (d) to assess the potential benefit of extending existing detection algorithms to accommodate the stochastic resonance effect for both visual and hyperspectral imagery, and (e) to outline the future research goals for algorithm development and implementation.