ADAPTIVE ENDMEMBER EXTRACTION BASED SPARSE NONNEGATIVE MATRIX
FACTORIZATION WITH SPATIAL LOCAL INFORMATION
Huali Li
1
, Shutao Li
1
*, Liangpei Zhang
2
1. College of Electrical and Information Engineering, Hunan University, P.R. Chin
2.State Key Laboratory of Information Engineering in Surveying, Mapping, and Remote Sensing,
Wuhan University, P. R. China
ABSTRACT
Hyperspectral Unmixing aims at getting the endmember
signature and their corresponding abundance maps from
highly mixed Hyperspectral image. Nonnegative Matrix
Factorization (NMF) is a widely used method for spectral
unmixing because it can obtain better performance while
there is no pure pixels in the hyperspectral image. However,
many methods based on nonnegative matrix factorization
seldom consider the spatial information both on local and
nonlocal. To combine the spatial and spectral information
together to improve the unmixing accuracy, an adaptive
endmember extraction based sparse nonnegative matrix
factorization with spatial local information (ASNMF) is
proposed in this paper. A superpixel segmentation is to
obtain many meaningful regions which are spectral similar
and spatial adjacent. Endmember is adaptively extracted on
each superpixel to generate endmember set. Initialing the
endmember set, ASNMF could adaptively obtain the final
endmembers with the sparse nonnegative matrix
factorization. Both the experiments on synthetic and real
scene images show the effectiveness of the proposed
method for hyperspectral unmixing.
Index Terms—adaptive endmember extraction, sparse
nonnegative matrix factorization, hyperspectral remote
sensing imagery
1. INTRODUCTION
Hyperspectral image has been widely applied in many area,
since it contains both high spectral resolution of
approximately 10 nm and plenty of spatial information.
Most related research assumes that each pixel vector
comprises the response of a single underlying material in
the scene. However, if the spatial resolution of the sensor is
low, different materials may jointly occupy a single pixel.
The resulting spectral measurement will be a mixed pixel,
composed of the individual pure spectra (i.e., endmembers)
and their corresponding fractional abundances.
According to the assumption as to whether or not pure
pixels exist in the image, endmember extraction methods
can be classified into two groups. One is the endmember
identify method, which assumes that the pure pixel existed
in the images, such as the vertex component analysis (VCA)
[1], and automatic morphological endmember extraction
(AMEE) [2], Piecewise Convex Multiple-Model
Endmember Detection (PCOMMEND) [3], etc. The other is
without this assumption, the endmember generation method
is introduced, which includes minimum volume constrained
non-negative matrix factorization (MVC-NMF) [4], L1/2
sparsity nonnegative matrix factorization (L1/2 NMF )[5],
Graph regularized nonnegative matrix factorization
(GNMF)[6], graph-regularized L1/2-NMF (GLNMF)
[7] ,etc.
To combine the spatial and spectral information
together, we proposed an adaptive endmember extraction
based sparse nonnegative matrix factorization with spatial
local information (ASNMF). ASNMF is consisted of
superpixel segmentation and sparse NMF. The superpixel
segmentation was used to obtain many meaningful regions
which are spectral similar and spatial adjacent. Instead of
the whole image, the sparse NMF is executed both on each
superpixel and the whole image, which can adaptively
obtain the endmember. So considering both local spatial and
spectral information, ASNMF could be adaptively obtain
the endmember.
The remainder of this paper is organized as follows.
Section II discusses the segmentation using superpixel and
presents the ASNMF algorithm. Section III describes the
experimental results with the simulated and real images.
Section IV provides the conclusions.
2. ASNMF
This section describes the proposed an adaptive
endmember extraction based sparse nonnegative matrix
factorization with spatial local information (ASNMF).
If the multiple scattering among distinct endmembers is
negligible and the surface is partitioned according to the
fractional abundances, linear mixed model can effectively
and mathematical simply to express. So linear spectral
unmixing is discussed in this paper.
1753978-1-4799-7929-5/15/$31.00 ©2015 IEEE IGARSS 2015