International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B4. Istanbul 2004
Comparing the spectral and texture classifications in table 2, we
see that spectral classification is better suited for those
landscape units with a specific spectral response pattern and
well differenciated from the rest of the units, such as pasture
land and cereal crops, or fallow. The distribution of grey levels
in these two classes is very homogeneous, so they are more
difficult to discriminate by texture methods. On the other hand,
texture techniques are very efficient in classifying lanscape
units that contain a high spectral heterogeneity, such as
scattered trees, forest-shrub and dense shrub. These classes are
not very accurate when classified using only spectral band.
Another interesting aspect is the the integration of spectral and
texture bands for classification has a synergic effect on the
résults, in some cases even improving the accuracy of both
groups of classes.
However, it is important to note that the reported results refer to
the inner areas of the texture units and not to the borders
between textures. In these areas, the border effect decreases the
overall accuracy to 47%. An example of this effect is shown on
the detail image of figure 6. Some previous tests have shown
how the post-processing operation, described for the energ
filters, increases the accuracy in the border areas in a 27% (Ruiz
et al., 2001)
Figure 6. Example of the border effect in texture classification.
4. CONCLUSIONS
This study has been focused on two main applications of texture
analysis in remote sensing: the classification of forest landscape
units and urban areas. The former holds a special interest for
mapping forest areas, the latter is a first step in monitoring
urban sprawl. Some important aspects can be concluded:
- The texture methods provide an alternative to the spectral
methods for the classification of forest units with a high
spectral heterogeneity, or when the classes are defined by
differences in vegetation density.
In urban classification, the texture methods are useful for
discriminating old urban areas and new residential spots,
but they introduce important errors in the classification of
industrial areas, so spectral information should be used in
addition to texture.
A universal criteria in order to use the idoneous texture
extraction method for classification does not seem to exist.
Therefore, the selection should be in funtion of the type of
landscape units defined in each application.
Furthermore, the combination of different texture methods
improves the classification results, especially when
combining statistical methods based on the GLCM with the
details of different levels obtained from the wavelet
transform. The Gabor filters allow an important part of the
texture information to be condensed into a few variables.
Before beginning the texture classification process, it is
important to previously select the methodological
parameters and features to reduce the volume of data and to
optimize the discrimination power of these techniques.
The main limitation for the standard application of texture
methods in image classification is probably the border
effect, inherent to texture analysis and which introduces
important errors in the transition areas between texture
units. Further work should be done to reduce this effect.
REFERENCES
Fatemi-Ghomi, N., Palmer, P.L., Petrou, M., 1996. Performance
of texture segmentation algorithms based on wavelets,
Technical Report. Electronic and Electrical Engineering
Department. University of Surrey.
Ferro, C.J. and Warner, T.A., 2002. Scale and texture in digital
image classification. Photogrammetric Engineering and Remote
Sensing, 68(1), pp. 51-63.
Haralick, R.M., K Shanmugam and Dinstein, 1973. Texture
features for image classification. IEEE Transactions on
Systems, Man. and Cibernetics. SMC-3 : pp. 610-622.
Hsiao, J.Y. and A.A. Sawchuk, 1989. Unsupervised image
segmentation using feature smoothing and probabilisatic
relaxation techniques. Computer Vision, Graphics and Image
Processing, vol. 48, pp. 1-21.
Laws, K.L, 1985. Goal-directed texture image segmentation.
Applications of Artificial Intelligence II, SPIE (548), pp.19-26.
Mallat, S.G., 1989. A theory of multirresolution signal
decomposition: The wavelet representation. ZEEE Trans. Patt.
Anal. Machine Intell., 11(7), pp. 674-693.
Ojala T, Mäenpää T, Pietikäinen M, Viertola J, Kyllônen J &
Huovinen S, 2002 Outex - New framework for empirical
evaluation of texture analysis algorithms. Proc. 16th
International Conference on Pattern Recognition, Quebec,
Canada, 1: pp. 701-706.
Randen, T. and Husoy, J.H., 1999. Filtering for texture
classification: A comparative study. IEEE Trans. Pal. Anal.
Machine Intell., 21(4), pp. 291-310.
Ruiz, L.A., Acosta, P., Fdez.-Sarria, A., Porres, M.J., Pardo,
LE. 2001. Mejora de la eficiencia en la segmentación 'de
imágenes por texturas mediante filtros de energía. Proc. of the
IX Spanish Congress on Remote Sensing, Lleida, pp. 477-480.
Sutton, R.N. and E.L. Hall, 1972. Texture measures for
automatic classification of pulmonary disease. IEEE
Transactions on Computers, vol. C-21, pp. 667-676.
Unser, M., 1995. Texture classification and segmentation using
wavelets frames. IEEE Trans. Image Processing, 4(11), pp.
1549-1560.
Van de Vower, G., Scheunders, P., Van Dyck, D. 1999.
Statistical Texture Characterization. from Discrete Wavelet
Representations. /EEE Trans. on Image Processing, 8(#), PP-
592-598.
ACKNOWLEDGMENTS
The authors wish to thank the financial support provided by the
Spanish Ministry of Science and Technology and the FEDER
(projects REN2003-04998 and BTE2002-04552), as well as ?
the Politechnic University of Valencia (project 2002-0627).
1114
of
us
Wt
int
fui
m«
rat
res
Soi
con
farı
lea
whi
AS
plu
site
can
AS!
furt
the
pre
volt
attei
volt
EPA
mod
Chir
eros
rain!
and
othe
regi
arid
1996
PSI/
