Advanced Methods for Representing and Processing Hyperspectral Image Data
Final rept. 1 Jul 2008-30 Jun 2011
CALIFORNIA UNIV IRVINE DEPT OF ELECTRICAL ENGINEERING AND COMPUTER SCIENCE
Pagination or Media Count:
We have extended our 3D spectralspatial Gabor representation to consider the effects of three-dimensional scene structure in hyperspectral images. We have shown that traditional spectralspatial models lead to ambiguities when classifying image regions due, in part, to changes that occur as the environmental conditions change. Our new models characterize the variation of vectors that are derived using spectralspatial features as the scene conditions change. We have shown that these models improve on the properties of standard techniques. The utility of this approach has been demonstrated using thousands of hyperspectral image regions that have been generated over a broad range of conditions. We have also modeled the effect on hyperspectral image sequences of ballistic impacts. We have shown that the multiband Gabor representation allows extraction of image properties that can be used to estimate properties of the impact. We have also applied the models to human skin in near-infrared images and demonstrated the potential for tissue classification.