=1,2,..., 14
OM P5 tr.
resolution
(11)
us4 x6 =
extract the
ig distance
1 T2 by us-
satures ex-
esolutions
ed, we can
H;da;ro DY
(12)
sed. The
Since the
olution on
with 10m
- to assure
lar for the
1er factors
. For the
h patch is
he size of
ze patches
the Build-
urban ar-
represent-
1s 202 im-
total, 670
ber of im-
fication.
lutions are
> scene on
y the con-
y different
| from the
eter p (see
y found to
extracting
ile param-
'arameters
] by using
directions
)m resolu-
ances 1, 2
the image
eatures of
Farm Building Vegetation
Figure 1: Examples of the SPOTS image patches taken on
Nanjing used for experiments(c)SPOTIMAGE. Top row: image
patches of the the classes with 2.5m resolution (512 x 512 pix-
els). Bottom row: image patches of the the classes with 10m
resolution (128 x 128 pixels).
4 EXPERIMENTS AND RESULTS
We combine the the radiometric features (Og), the GLCM fea-
tures (Og) and the Gaussian wavelet features (Ow) for the re-
trieval of the SPOT images at different reslutions.
In order to validate this combination of features, we have evalu-
ated the classification performance of different combinations of
the Gabor features, the GLCM features, the shape features and
the radiometric features (see (Luo et al., n.d.) for details of these
features). The results are shown in Figure 2, where the images of
10m resolution (Figure 2(a)) and of 2.5 resolution (Figure 2(b))
are randomly selected as training samples. The percentage of the
training samples vary from 10% to 60%. The test sets include all
the images of 10m and 2.5m resolutions. It can be seen that the
combination of the radiometric features (Or), the GLCM fea-
tures (© i) and the Gaussian wavelet features (©) can give the
best results.
For each retrieval, a key image patch (with 2.5m or 10m reso-
lution) is selected from from the database for the request. Its
radiometric, GLCM and Gasussian scale-space features are then
compared with the features of the other image patches in the
database. The Euclidean distance between the features are com-
puted as similarity measurement. The most similar image patches
are selected as the retrieval result. For each retrieval, 47 image
patches are shown. In Figures 3 to 5, the results of three retrieval
experiments are shown. The three key image patches belong to
the three different classes: Building, Farm and Vegetation.
Several remarks can be drawn from the retrieval results:
e Although the visual appearences (the contrasts) of the im-
ages with 2.5m resolution and the images with 10m resolu-
tion are different, the retrieval results are quite good. For
the Farm and the Building classes, all the retrieved images
belong to the same class of the key image. While for the
Building and the Vegetation class, there is only one retrieved
image belongs to a different class to the key image. This in-
dicates that though the radiometric features are usually perti-
nent, the other two feature sets (the GLCM and the Gaussian
wavelet features) are also important. Moreover, the compar-
ison schemes for the features extracted from images at dif-
ferent reslutions proposed in Section 2 are accurate enough
for the joint retrieval.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B3, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
different features, 10m training, result of all
1 L
2 0.9
I
2 8
e t
o 0.8
8 j
© | H
207} 9,9, HO
s R'UHTTW
—e— 849, tO, tO.
0.6r © +0, +O
a : : ge ig) ore BASSE,
20 10% 20% 30% 40% 50% 60% 70%
different Training samples
(a)
different features, 2.5mtraining, result of all
1 L
- 09.
Lt
2
ë
o 0.8
o
o
&
20.7;
AX
0.6
0.5
0 10% 20% 30% 40% 50% 60% 70%
different Training samples
(b)
Figure 2: Classification results of different feature sets (a) when
the images of 10m resolution is selected as training samples; and
(b) when the images of 10m resolution is selected as training sam-
ples.