HYPERSPECTRAL IMAGE CLASSIFICATION BASED ON SPECTRAL AND

GEOMETRICAL FEATURES
Bin Luo and Jocelyn Chanussot
GIPSA-Lab, 961 rue de la Houille Blanche, 38402 Grenoble, France.
([email protected])
Gocelyn. [email protected])

ABSTRACT

In this paper, we propose to integrate geometrical features,

such as the characteristic scales of structures, with spectral
features for the classification of hyperspectral images. The
spectral features which only describe the material of structures can not distinguish objects made by the same material
but with different semantic meanings (such as the roofs of
some buildings and the roads). The use of geometrical features is therefore necessary. Moreover, since the dimension of
a hyperspectral image is usually very high, we use linear unmixing algorithm to extract the endmemebers and their abundance maps in order to represent compactly the spectral information. Afterwards, with the help of these abundance maps,
we propose a method based on topographic map of images to
estimate local scales of structures in hyperspectral images.
The experiment shows that the geometrical features can
improve the classification results, especially for the classes
made by the same material but with different semantic meanings. When compared to the traditional contextual features
(such as morphological profiles), the local scale feature provides satisfactory results without considerably increasing the
feature dimension.
1. INTRODUCTION

The classification of hyperspectral remote sensing images is

a challenging task, since the data dimension is considerable
for traditional classification algorithms, typically several hundreds of spectral bands are acquired for each image. These
spectral bands can provide very rich spectral information of
each pixel in order to identify the material of the objects.
However, the spectral information alone sometimes does not
allow the separation of structures. For example, the roofs of
some buidings and the roads can be made by the same material (alsphalt). Therefore, contextual information, geometrical features for example, is necessary for classification task
of hyper spectral images.
The major difficulty to calculate the contextual information ofhyper spectral images is the high dimension ofthe data.
This work is funded by French ANR project VAHINE.

978-1-4244-4948-4/09/$25.00 2009 IEEE

To reduce the data dimension, both supervised and non supervised methods are proposed. The supervised methods, such as
band selection [1] [2], Decision Boundary feature extraction
and Non-Weighted feature extraction [3], transform the data
according to the training set in order to improve the separability of the data. However, the supervised methods depend on
the quality of training set. The unsupervised methods, such
as PCA (principle component analysis) or ICA (independant
component analysis), optimise some statistical criteria (such
as the most un correlated components or the most independant
components) to project the data onto a sub space with lower
dimension. The application of these methods on hyperspectral data can be found in [4] [5] [6]. However, the components
obtained by optimising the statistical criteria do not necessarily have physical meanings. In this paper, we propose to
use Vertex Component Analysis (VCA) to reduce the dimension of hyperspectral data [7]. We suppose that the spectrum
of each pixel is a linear mixture of the spectra of different
chemical species (refered as endmembers). This linear mixture model is physically valid for the reflectance ofthe surface
of the Earth without being affected by the aerosol. The VCA
can separate the spectra of these endmembers and estimate
their spatial abundances. Since the number of the endmembers is much less than the number of spectral bands, these
abundance maps can be considered as a compact representation of spectral information provided by the hyperspectral
image.
In order to extract the contextual information from remote
sensing images, one can find the methods based on Markov
Random Field (MRF) [8]. However, the MRF based methods provide only statistical information on the neighborhood
of the considered pixel. Another family of methods for contextual information extraction are based on morpholigical operators, which allow to extract descriptive features, such as
the geometrical features about the structures [9]. [4] [5] extract the extended morphological profiles (EMP) ofthe principle components ofhyper spectral images for the classification.
However, since all morphological methods require a structred
element, the features extracted by such methods depend on
the used structured element. Moreover, EMP increase con-

2. LINEAR UNMIXING OF HYPERSPECTRAL DATA

Hyperspe ctrallmage
Linear Unmixing
(Section 2)

Segmentation
(Section 3)

Scale extraction
(Section 3)

Classification by SVM

Fig. 1. Scheme of the paper.

siderably the feature dimension. For example, in order to describe the geometrical information of one pixel on a principle
component, the EMP method in [4] [5] uses 8 values. In [10],
based on topographic map of gray scale image, the authors
define a characteristic scale of each pixel in panchromatic remote sensing images. The main idea is that for each pixel , the
contrasted structure containing it is extracted, and the scale
of this structure is defined as the scale of this pixel. In this
paper, we try to extend the algorithm presented in [10] for
extracting a local characteristic scale for each pixel on hyperspectral images with the help of the abundance maps obtained
by VCA. The main advantages of our proposed method when
compared to EMP are two-fold. At one hand, no structured
element is required for this method. Secondly, we use only
one value (the local scale) to describe the geometrical feature
of a spatial position rather than the EMP which need many
values.
The plan of this article can be illustrated by in Figure I.
In Section 2, we introduce very briefly the Vertex Component
Analysis (VCA) for estimating the abundance maps of endmembers in order to reduce the dimension of hyperspectral
data. In Section 3, we introduce the method presented in [10]
for extracting the local scale of panchromatic images and extend this method to hyperspectral images. In Section 4, we
classify a hyperspectral image by using both the spectral features and the scale feature in order to show the effeciency of
the local scale feature.

We note X the matrix representing the hyperspectral image cube , where X = {X I ,X2 , ... ,X N a } and
Xk =
T,
{ XI ,k , X2 ,k , . . . ,xN,, ,d
Xl ,k is the value of the kth pixel at
the lth band. We assume that the spectrum of each pixel is a
linear mixture of the spectra of N; endmembers, leading to
the following model:
X = MS +n

(I)

where M = {m- , m2, ... , mNJ is the mixing matrix, where

m., denotes the spectral signature of the nth endmember.
S = { 5 I , 52 , .. . , 5 N o } T is the abundance matrix where
5 n = {Sn ,l , S n ,2 , . . . , S n ,N a } (S n ,k E [0, 1] is the abundance of the nth endmember at the kth pixel). n stands for
the additive noise of the image. In [7], the Vertex Component Analysis (VCA) is proposed as an effecient method for
extracting the endmembers which are linearly mixed. The
main idea is to extract the vertex of the simplex formed by
M which contains all the data vectors in X. The sum of the
abundances of different endmembers at each pixel is one, i.e,
L:~ l S n ,k = 1, which is called the sum-fa-one condition.
Therefore the data vectors Xl are always inside a simplex of
which the vertex are the spectra of the endmembers. VCA iteratively projects the data onto the direction orthogonal to the
subspace spanned by the already determined endmembers.
And the extreme of this projection is the new endmember
signature. The algorithm stops when all the p endmembers
are extracted, where p is the number of endmembers. Even
though the number of endmembers is much smaller than the
number of spectral bands, i.e, p << N s , abundance maps of
the p endmembers can provide the same spectral information
as the hyperspectral image with N; spectral bands. Therefore
this linear unmixing step can be considered as a dimension
reduction of data. Moreover, the values on the abundance
maps for one given pixel which represent the proportions
of different chemical species on this pixel are comparable,
which is essential for estimating the characteristic scale of
this pixel. On the contrary, values of one pixel obtained by
other dimension reduction methods (such as PCA or ICA)
are not comparable. In the next section, we can see that the
comparability of the proportions of different components obtained by VCA is very important for extracting the local scale
of a structure.

3. LOCAL CHARACTERISTIC SCALE OF

HYPERSPECTRALIMAGE
In [10], the authors propose a method based on the topographic map of the image to estimate the local scale of each
pixel in the case of gray scale remote sensing images. The
idea is that , for each pixel, the most contrasted shape containing it is extracted, and the scale of this shape defines
as the characteristic scale of this pixel. The topographic

map [11], which can be obtained by Fast Level Set Transformation (FLST) [12], represents an image by an inclusion
tree of the shapes (which are defined as the connected components of level sets) . An example of such inclusion tree is
shown in Figure 2. For each pixel (x, y), there is a branch
of shapes f i(X, y) (fi-l C f i) containing it. Note l(Ji) the
gray level of the shape f i(X, y), S(Ji) its area and P(Ji)
its perimeter. The contrast of the shape f i(X, y) is defined
as C(Ji) = II (Ji+ l ) - I( Ji) I. The most contrasted shape
Ji(x , y) of a given pixel is defined as the shape containing
this pixel, of which the contrast is the most important, i.e.

(a)
A

Ji(x , y)

ar gmax{C(Jj(x , y)}

:5255

(2)

J
C ,,40

The scale of this pixel E (x , y) is defined as the area of the

most contrasted shape divided by its perimeter, i.e.

E (x , y)

S(Ji(x , y)) j P( Ji(x , y)) .

2;128

H :50

(b)
Fig. 2. Example ofFLST: (a) Synthetic image; (b) Inclusion
tree obtained with FLST.
object is mainly made by the nth endm ember, the scale values
calculated on the abundance maps of the other endmembers
for this object have no meaning. Therefore, we try to define
one single scale value for each pixel which corresponds to
the scale of this pixel on the most significant abundance map.
More precisely, since the values of different abundance maps
are comparable, the characteristic scale at the spatial position
(x, y) for a hyperspectral image is defined as

E (x , y)
L (k ,l )E! ; (X,y ) l( k, l)

S(Ji(x, y))

G 90

F 2:150

(3)

Since the optical instruments always blur remote sensing images, several shapes with very low contrasts can belong to
the same structure. In order to deal with the blur, the authors of [10] propose a geometrical criterion to cumulate the
contrasts of the shapes corresponding to one given structure.
The idea is that the difference of the areas of two succesive
shapes (for example f i and fi+l ) corresponding to a same
structure is proportional to the perimeter ofthe smaller shape ,
i.e. S(Ji+l) - S(Ji) rv >"P(Ji), where>" is a constant. It is
shown in [10] that this local scale corresponds very well to
the size of a structure and it is a very significant feature for
characterizing a structure in remote sensing images. Remark
that the most contrasted shapes extracted in an image form a
partition of this image. We can therefore have a simplified
image by defining the value of each pixel as the mean value
of the gray levels of the pixels in its most contrasted shape,
i.e. the value of (x, y) in the simplified image is

l(x y) _
,-

D :5128

(4)

where I( k , l) is the gray level of the pixel (k , l) in the original image. This value is more significant for characterizing
spectrally the pixel than its original gray level since it takes
into account the context pixels.
In order to extend the estimation of local scale to hyperspectral images , the scales on all the abundance maps of the
p endmembers obtained by veA are first computed. Therefore, for each spatial position (x, y) in an hyperspectral image, there are p scale values. For the pixel (x, y) on the nth
abundance map, we note En (x, y) its local scale calculated by
Equation (3), Ji,n (x , y) the most contrasted shape extracted
by Equation (2), and In( x , y) its value of the simplified image defined by Equation (4). The simplest way to use the scale
features ofa pixel is to use all the En(x, y) values as features
for classification. However, the major drawback is that if an

Eft(x , y)

(5)

where n = arg maxn {C (Jin (x , y))}.

The feature vector 8(x, 'y) ofa given pixel (x, y) used for
classification contains the values of the simplified image of
all the abundance maps and its local scale defined by Equation (5), i.e.

8(x ,y)

{ll( x ,y) , ... I Nc(x ,y) , E( x ,y)} .

(6)

4. EXPERIMENT
In this section , we use the features defined by Equation (6)
to classify a hyperspectral image taken by the instrument ROSIS (Reflective Optics System Imaging Spectrometer) over
the University of Pavia, Italy (see Figure 3(a)). The image (with a spatial resolution of 1.3m) contains 340 x 610
pixels and 103 spectral bands covering visible and near infrared light. The image is manually classified into 9 classes

and the definitions of these classes are shown in Figure 4.

For classification purpose, the training set contains 3921 pixels (see Figure 3(b)) while the test set contains 42776 pixels (see Figure 3(c)). Since in this image, there are mainly
3 endmembers present (vegetation , bare soil and metal root),
we extract 3 endmembers and their abundance maps by using VCA (see Section 2), which are shown in Figures 5(a)(c). Afterwards we compute the simplified images of these
abundance maps by using Equation (4). The simplified images are shown in Figure 5(d)-(t). According to Equation (5),
we estimate the local scale for each pixel of this image. The
scale map is shown in Figure 6(a). Therefore the feature vector 8(x , y) ofa pixel (x , y) defined by Equation (6) contains
only 4 values: 3 values of the simplified abundance maps
(ll( X, y), 12(x , y), l a(x , y)) and one scale value (E( x , y)).
We have classified the image by using the original hyperspectral data (all 103 bands) and the feature vector 8 defined by Equation (6). Kernel methods are proved to be effecient algorithms for hyperspectral image classification. [13]
Therefore, we use the Support Vector Machine (SVM) with
Gaussian kernel as classifier. The optimal scale parameter of
Gaussian kernel is selected by 5-fold cross validation on the
training set.
The overall accuracies and the classification accuracies of
each class for both cases are shown in Table 1. In Figure 6(b)
and (c), the classification results obtained on the whole image is shown. In [5], the authors proposed to use PCA and
Kernel PCA (KPCA) to reduce the dimension of hyperspectral images. Extended Morphological Profiles (EMP) are then
extracted from the principle components obtained by PCA (3
principle components are extracted) or KPCA (12 principle
components are extracted) as features for classifying the hyperspectral image. In Table 1, we show the classification results obtained in [5] on the same data set by using SVM with
Gaussian kernel.
It can be seen that by using the features proposed in our
article, the classification results improve considerably when
compared to the results obtained on original hyperspectral
data. Recall that the length of the feature vector 8 for each
pixel is only 4 which is much less than the number of spectral
bands (103). Moreover, the results obtained by using PCA
and EMP are very similar with the results obtained by the
features proposed , since we have previously mentioned that
the major information described by EMP on a pixel is its local scale. Indeed, the local scale of a pixel can be considered
as the width of the morphological filter by which the absolute
differential EMP value reaches its maximum. However, the
number of features extracted by PCA and EMP is 27, which
is much larger than the number of features in 8. The results
obtained by using KPCA and EMP are better than the results
obtained by using 8 and the results by using PCA and EMP.
However, the feature dimension, which is 108, is even higher
than the dimension of original data (10 3). And the improvement of classification accuracies is mainly concentrated on

(b)

(a)

(c)

Fig. 3. (a) Image of Pavia University (R-band 90, G-band 60,

B-band 40); (b) Training set; (d) Test set.
I

CL~ SSES

Poin ts in T ra ining S.... t

Trees

524

As phalt
B itume n

,~ 8

375
39 2
26 5

~ 31

Grave l
(p ainted ) metal sheets
Shadow

Self-Blocking Bricks

, l~

M ead ows
Ba re So il

53 2

,~ o

The matic Colour

Fig. 4. Definitions of the classes.

the class of Meadows, which is not well defined in the ground
truth.
It has to be remarked that by using only the spectral information (i.e. the spectrum of each pixel), it is very difficult
to distinguish the class Asphalt and Bitumen, since they are
made by the same material. According to the ground truth
of Figure 3(c), the only differen ce is that Asphalt is used for
the roads while Bitumen is used for the building roofs. However, it can be seen from Figure 6(a) that the scale values of
the pixels on the building roofs are very different from the
scales of the pixels on the roads. Therefore the classification
results of these two classes by using the feature vectors 8 are
much more accurate. This phenomenon can be illustrated by
Figure 6(d)-(t), where we have zoomed the classification results obtained by using original data and the feature vectors
8 on a building roof made by bitumen. It can be seen that
the building roof and the road are classified as the same class
in Figure 6(e). In contrary, by adding the local scale feature,
these two structures are well separated.
5. CONCLUSION

In this article, we have proposed to integrate geometrical feature, the characteristic scales of structures, for the classification ofhyperspectral images. In order to reduce the dimension
ofthe data, we used a linear unmixing algorithm to extract the

Feature
Number of features

Original data
103

Feature vector e
4

EMPpCA[5j
27

EMPKPCA[5j
108

Overall accuracy
Tree
Asphalt
Bitumen
Gravel
(painted) metal sheets
Shadow
Self-Blocking Bricks
Meadows
Bare Soil

76 .01 %
98 .59 %
78.16 %
89 .02 %
64 .55 %
99.47%
99 .89 %
91.01 %
64 .23 %
82 .72 %

91.54%
94 .52 %
96 .27 %
99 .32 %
84 .61 %
99 .55 %
96 .30 %
99.70%
85 .80 %
96 .56 %

92 .04 %
99 .22 %
94 .60 %
98 .87 %
73 .13 %
99 .55 %
90 .07 %
99 .10 %
88 .79 %
95 .23 %

96 .55 %
99 .35 %
96 .23 %
99 .10 %
83 .66 %
99 .48 %
98 .31 %
99.46%
97 .58 %
92 .88 %

Table 1. Classification results

(a)

(d)

(b)

(e)

(c)

(f)

Fig. 5. (a)-(c) Abundance maps of the 3 endmembers extracted by VCA ; (d)-(f) the simplified images of these abundance maps.

(a)

(b)

(d)

(e)

(f)

Fig. 6. (a) Scale map of the hyperspectral image, bright pixel

corresponds to big scale while dark pixel corresponds to small
scale; (b) Supervised classification results obtained on original data set; (c) Supervised classification results obtained
on the feature set obtained by Equation (6). (d)-(f): Zoom
around a building on the: (d) original image; (e) classification results of (b) ; (f) classification results of (c).

endmembers and their abundance maps contained in a hyperspectral image. The abundance maps can be considered as a
compact representation of spectral information of this image,
since the number of endmembers contained in a hyperspectral image is much smaller than the number of spectral bands.
With the help ofthese abundance maps, we extend the method
proposed in [10] to hyperspectral images to estimate the characteristic scales of the structures. The experiments show that
the use of the scale feature, the classification results improve
considerably, especially for the objects made by the same material but with different semantic meanings. By using the features proposed, the classification results are very similar to
the results obtained by using the methods based on PCA and
EMP, even though the number of features proposed is much
less.
Acknowledgement The authors would like to thank Pro
Paolo Gamba, University of Pavia, for providing the data and
the ground truth of the classification.
6. REFERENCES

[1] S. B. Serpico and L. Bruzzone, "A new search algorithm

for feature selection in hyperspectralremote sensing images," IEEE Trans. on Geoscience and Remote Sensing,
vol. 39, no. 7, pp. 1360 - 1367, July 2001.
[2] B. Guo, S.R. Gunn, R. I. Damper, and J.D.B. Nelson,
"Band selection for hyperspectral image classification
using mutual information," IEEE Geoscience and Remote Sensing Letters, vol. 3, no. 4, pp. 522-526, 2006.
[3] D.A. Landgrebe, Signal Theory Methods in Multispectral Remote Sensing, John Wiley and Sons, New Jersey,
2003.
[4] M. Fauvel, J. A. Benediktsson, J. Chanussot, and J. R.
Sveinsson, "Spectral and spatial classification of hyperspectral data using SVMs and morphological profile,"
IEEE Transaction on Geoscience and Remote Sensing,
vol. 46, no. 11, November 2008.
[5] M. Fauvel, J. Chanussot, and J. A. Benediktsson, "Kernel principal component analysis for the classification
of hyperspectral remote-sensing data over urban areas,"
EURASIP Journal on Advances in Signal Processing,
2009, to appear.
[6] J.A. Palmason, J.A. Benediktsson, J.R. Sveinsson, and
J. Chanussot, "Classification of hyper spectral data from
urban areas using morpholgical preprocessing and independent component analysis," in IEEE IGARSS '05 International Geoscience and Remote Sensing Symposium, Seoul Korea, 2005, pp. 176-179.
[7] J. M. P. Nascimento and J. M. B. Dias, "Vertex component analysis: A fast algorithm to unmix hyperspectral

data," IEEE Trans. Geoscience and Remote Sensing,

vol. 43, no. 4, pp. 898 - 910, April 2005.
[8] G. Poggi, G. Scarpa, and J. Zerubia, "Supervised segmentation of remote sensing images based on a treestructure MRF model," IEEE Trans. on Geoscience and
Remote Sensing, vol. 43, no. 8, pp. 1901-1911,2005.
[9] M. Pesaresi and J. A. Benediktsson, "A new approach
for the morphological segmentation of high-resolution
satellite imagery," IEEE Trans. on Geoscience and Remote Sensing, vol. 39, no. 2, pp. 309-320, Febrary 2001.
[10] B. Luo, J.-F. Aujol, and Y. Gousseau, "Local scale measure from the topographic map and application to remote
sensing images," SIAM Multiscale Modeling and Simulation, to appear.
[11] V. Caselles, B. ColI, and J.-M. Morel, "Topographic
maps and local contrast changes in natural images," Int.
J Compo Vision, vol. 33, no. 1, pp. 5-27,1999.
[12] P. Monasse and F. Guichard, "Fast computation of a
contrast-invariant image representation," IEEE Trans.
on Image Processing, vol. 9, no. 5, pp. 860-872, may
2000.
[13] G. Camps-Valls and L. Bruzzone, "Kernel-based methods for hyperspectral image classification," IEEE Trans.
on Geoscience and Remote Sensing, vol. 43, no. 6, pp.
1351-1362,2005.