This SBIR will examine the use of hyperspectral mixture models that may be used to create shadow invariant images for material matching and hidden target detection and tracking. This proposal introduces an "Illumination-based Linear Unmixing Procedure" (ILUMP). ILUMP assumes that every image is illuminated by two spectrally different illumination sources and that at each pixel the illumination source may be partly occluded. This is typically the case in outdoor daytime scenes where there is both sunlight and skylight illumination. The ILUMP formulation requires more sophisticated and computationally expensive model estimation. But this more precise model helps improve the results of applications such as shadow removal and material matching. The proposed contract would further the development of this model by developing model parameter estimation methods and applications of the proposed technologies.
Keywords: Hyperspectral Image Processing , Unmixing
