A probabilistic approach to spectral unmixing

Authors:
Cong Phuoc Huynh;Antonio Robles-Kelly
Affiliations:
School of Engineering, Australian National University, Canberra, ACT, Australia;School of Engineering, Australian National University, Canberra, ACT, Australia and National ICT Australia, Canberra, ACT, Australia
Venue:
SSPR&SPR'10 Proceedings of the 2010 joint IAPR international conference on Structural, syntactic, and statistical pattern recognition
Year:
2010

Citing 7
Cited 0

A physical approach to color image understanding

International Journal of Computer Vision
A generative model for separating illumination and reflectance from images

The Journal of Machine Learning Research
Specular Free Spectral Imaging Using Orthogonal Subspace Projection

ICPR '06 Proceedings of the 18th International Conference on Pattern Recognition - Volume 01
Some Unmixing Problems and Algorithms in Spectroscopy and Hyperspectral Imaging

AIPR '06 Proceedings of the 35th Applied Imagery and Pattern Recognition Workshop
Illumination Chromaticity Estimation Based on Dichromatic Reflection Model and Imperfect Segmentation

Computational Color Imaging
A comparison of computational color constancy Algorithms. II. Experiments with image data

IEEE Transactions on Image Processing
Edge-Based Color Constancy

IEEE Transactions on Image Processing

Quantified Score

Hi-index	0.00

Visualization

Abstract

In this paper, we present a statistical approach to spectral unmixing with unknown endmember spectra and unknown illuminant power spectrum. The method presented here is quite general in nature, being applicable to settings in which sub-pixel information is required. The method is formulated as a simultaneous process of illuminant power spectrum prediction and basis material reflectance decomposition via a statistical approach based upon deterministic annealing and the maximum entropy principle. As a result, the method presented here is related to soft clustering tasks with a strategy for avoiding local minima. Furthermore, the final endmembers depend on the similarity between pixel reflectance spectra. Hence, the method does not require a preset number of material clusters or spectral signatures as input. We show the utility of our method on trichromatic and hyperspectral imagery and compare our results to those yielded by alternatives elsewhere in the literature.