Output details

A fast separability-based feature selection method for high-dimensional remotely-sensed image classification

Significance of output:

<24>This paper presents a new approach to address the curse of dimensionality in high-dimensional remotely sensed data classification. A theoretical approach is developed showing how a mutual information criterion can be decomposed into low dimensional sub-problems, providing a very efficient implementation. The work was successfully evaluated using publically available image data from the AVIRIS hyper-spectral sensor providing state-of-the-art results and has influenced the work of others in the hyper-spectral community.