Spectral unmixing using nonnegative tensor factorization

Authors:
Qiang Zhang;Han Wang;Robert Plemmons;V. Paul Pauca
Affiliations:
Wake Forest University, Winston-Salem, NC;Wake Forest University, Winston Salem, NC;Wake Forest University, Winston Salem, NC;Wake Forest University, Winston Salem, NC
Venue:
ACM-SE 45 Proceedings of the 45th annual southeast regional conference
Year:
2007

Citing 0
Cited 1

Multidimensional Noise Removal Based on Fourth Order Cumulants

ACIVS '08 Proceedings of the 10th International Conference on Advanced Concepts for Intelligent Vision Systems

Quantified Score

Hi-index	0.00

Visualization

Abstract

Three major objectives in processing hyperspectral image data of an object (target) are data compression, spectral signature identification of constituent materials, and determination of their corresponding fractional abundances. Here we propose a novel approach to processing hyperspectral data using nonnegative tensor factorization (NTF), which reduces a large tensor into three factor matrices, the Khatri-Rao product which approximates the original tensor. This approach preserves physical characteristics of the data such as nonnegativity and can be used to satisfy all three major objectives. Test results are reported for space object identification.