Material Signature Orthonormal Mapping in Hyperspectral Unmixing to Address Endmember Variability
Year: 2016
Type: article
Abstract: A new hyperspectral unmixing algorithm which considers endmember variability is presented. In the proposed algorithm, the endmembers are represented by correlated random vectors using the stochastic mixing model. Currently, there is no published theory for selecting the appropriate distribution for endmembers. The proposed algorithm first uses a linear transformation called material signature orthonormal mapping (MSOM), which transforms the endmembers to correlated Gaussian random vectors. The MSOM transformation reduces computational requirements by reducing the dimension and improves discrimination of endmembers by orthonormalizing the endmember mean vectors. In the original spectral space, the automated endmember bundles (AEB) method extracts a set of spectra (endmember set) for each material. The mean vector and covariance matrix of each endmember estimated directly from endmember sets in the MSOM space. Second, a new maximum likelihood method, called NCM_ML, is proposed which estimates abundances in the MSOM space using the normal compositional model (NCM). The proposed algorithm is evaluated and compared with other state-of-the-art unmixing algorithms using simulated and real hyperspectral images. Experimental results demonstrate that the proposed unmixing algorithm can unmix pixels composed of similar endmembers in hyperspectral images in the presence of spectral variability more accurately than previous methods.
Authors: Ali Jafari, Mohammad Ebadzadeh, Reza Safabakhsh
Institution: Amirkabir University of Technology
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $330
OpenAlex ID: W4240901418