In: Wagner W„ Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part 7B 
In: Wagn 
3.3 Texture features Behaviour 
Entropy is negatively correlated with window size when tested 
separately. However, when combined with the spectral bands 
tends to increase as the window size raises. The contrast is 
correlated positively with window size in both, when tested 
separately and when combined with spectral bands. In turn, the 
ASM presents negative correlation in both situations. 
This indicates that the second angular momentum may play an 
important role when the window size varies between the values 
11 and 30. Unlike the contrast that reaches the best results with 
windows between 30 and 45. Entropy presents a peculiar 
behaviour because it presents very significant results in all sizes 
of windows, but these tend to improve as the window size 
increases. In the Entropy plus Spectral graph, we can observe 
two peaks: one for windows of size 27 and one for windows of 
size higher than the 39. 
№**»»*<*« 
(c) ASM+Spectral (d) ASM 
Window 
Cont 
+S 
ASM 
+S 
IDM 
+S 
Ent 
+S 
Corr 
+S 
Window 
1,00 
Cont+S 
0,89 
1,00 
ASM+S 
-0,83 
-0,87 
1,00 
IDM+S 
0,38 
0,08 
0,13 
1,00 
Ent+S 
0,86 
0,79 
-0,66 
0,27 
1,00 
Corr+S 
-0,60 
-0,58 
0,75 
0,29 
-0,46 
1,00 
Table 5. Correlation Matrix of all classification results using 
each texture feature separately combined with spectral band, a 
fixed lag distances of 3 and odd values of window sizes 
between 11 and 45. 
Window 
Cont 
ASM 
IDM 
Ent 
Corr 
Window 
1,00 
Cont 
0,86 
1,00 
ASM 
-0,69 
-0,61 
1,00 
IDM 
-0,71 
-0,74 
0,84 
1,00 
Ent 
-0,63 
-0,69 
0,44 
0,65 
1,00 
Corr 
-0,94 
-0,69 
0,67 
0,63 
0,44 
1,00 
Wttefctwbn 
(e) IDM+Spectral 
(f) IDM 
(I) Corr.+Spectral 
(J)Corr. 
Figure 4. Dispersion Graphs of classification results of texture 
features according to window size for each feature processed 
combined and not combined with spectral bands. 
Table 6. Correlation Matrix of all classification results using 
each texture feature separetely, a fixed lag distances of 3 and 
odd values of window sizes between 11 and 45. 
4. CONCLUSION 
(a) Cont.+Spectral (b) Cont. 
The red band of the IKONOS image proved to be suitable for 
extraction of texture information to be used in the classification 
of aquatic vegetation in high-resolution images. Smaller 
distances were more efficient for image classification, as well 
as window sizes larger fared better. The use of a single texture 
feature combined with the spectral bands was very 
efficient. Among the parameters of texture Entropy combined 
with spectral bands produced good results for classification in 
all window sizes. While the contrast obtained the best results in 
windows of larger sizes and Second Angular Momentum 
showed good results in the windows of smaller sizes. 
The auth 
(http://w\ 
image ar 
Fonseca ; 
also than 
Universic 
Marco Ol 
Biodivers 
para 
atlas 
Biodi 
Barbosa, 
wetla 
pande 
Canai 
Clausi, D 
Gabor an 
ice in 
Davis, P. 
Evalu 
vegeti 
Geolc 
Dechka J 
W.Inj 
vegeti 
image 
Remo 
Franklin, 
Mana 
Haralick, 
textur 
Engin 
Maillard, 
throuj 
Remo 
Maillard 1 
The 
planni 
Melack, J 
Remot 
Envin 
Sons, 
Mitsch, 
John I 
National 
Wetla 
Water 
Ramsar C 
Manui 
66