^ ^ ^ ^ 
scales of window 
-S— bandi 
— band2 
-£— band3 
band4 
-•— band5 
H— band7 
-bandi - mean the combination of band51-bandi-band54 
Figure 3. Relationship between the OIF value and the scales 
of the window 6 
6. CONCLUSION 
Combine discussing the OIF value with the scales of sliding 
window, which make the final results more extensive meaning 
in different scales. Because of different sizes represent 
different scales in sliding window method, the OIF value which 
calculated represent the distribution situation of different 
scales. 
[1] Manik Varma,2004, .Unifying statistical texture 
classification frameworks. Image and Vision Computing, 
vol(22).pp:l 175-1183 
[2] M.Unser, 1999,Multiresolution Feature Extraction and 
Selection for Texture Segmentation. Pattern Analysis 
and Machine Intelligence. Vol.21(4) 
[3] Bea Thai, 1999 Spatial Filter Selection Selection for 
Illumnation-Invariant Color Texture Discrimination. 
Computer Vision and Pattern Recognition-Volume 2 06 
23-06, Fort Collins, colorade. 
[4] S.E.Grigorescu,N,Petkov, and P.Kruizinga,2000,A 
Comparative Study of Filter Based Texture Operators 
Using Mahalanobis Distance International conference on 
Pattern Recognition-Volume 3 09 03 - 09, 
Barcelona,Spain 
[5] H.Cheng and C.A.Bouman,2001,Multiscale Bayesian 
segmentation using a trainable context modelJEEE 
Transaction on Image Processing, 10(4):511-525 
[6] DeWouwer G V, Scheunders P ,Dyck D 
V,2001 Statistical texture characterization from wavelet 
representations./£'£'£' Singal Processing Letter, 
8(5): 125-128 
[7] G.Val de Wouwer, 
P.Scheunders,D.Van.Dyck, 1999,Statistical texture 
characterization from wavelet representation, IEEE 
Traction on Inamge Processing, 8(5):592-598 
[8] Novak I D,Soulakellis N,2000.Identifying geomorphic 
features using LANDSAT-5/TM data processing 
techniques on 
Lesvos, Greece. Geomorphology, 34:101-109 
REFERENCE 
The author is grateful for support through the following 
Grants: project of central scientific research institution with 
public welfare (YWF200809 and YWF0721/AQ01 , 
YWF0723/AQ03) 
ACKNOWLEDGEMENT 
It can be concluded the following points by contrasting display 
effect of many images and referencing Figure 3 relation 
between the OIF value and the scales of the window. 
(l)the optimal sliding window is 3x3 . The origin dada in 
this paper is ETM+ RS image whose ground resolution is 30 
meters. So 3 pixels in the image mean 90 meters in the actual 
length. In this scale, most of medium and large-scale geological 
phenomena can be recognized easily in the image. And small 
geological phenomena will be filtered which will decrease the 
disturbance from the theme. In this scale of window, RS image 
has the better effect than that of any other else scales. The more 
large of the scale is , the worse of display effect will be. 
The International Archives of the Photogrammetry, Remote Sensing 
calculated by above methods we have talked. Then the OIF 
value we have got can be statistical analysis for finding the 
optimal sliding window. 
The concrete contents are as follows: 
\SfPX 
size=3 
size=5 
size=7 
size=9 
-bandl- 
316.4495 
229.9329 
198.9636 
181.0253 
-band2- 
431.3088 
286.9773 
237.6239 
211.1394 
-band3- 
377.6173 
285.1752 
251.4869 
231.3289 
-band4- 
128.9929 
126.7493 
125.1496 
122.9298 
-band5- 
142.4992 
140.998 
139.3708 
136.8546 
-band7- 
192.24 
190.5205 
187.5389 
183.2421 
size=l 1 
size=13 
size=15 
size=17 
-bandi- 
169.5374 
164.2567 
160.0913 
155.4631 
-band2- 
194.6742 
186.542 
180.3785 
174.1038 
-band3- 
217.9556 
212.1879 
206.9307 
199.9323 
-band4- 
121.3445 
122.3527 
123.1257 
122.8622 
-band5- 
134.9143 
135.7672 
136.3186 
135.7962 
-band7- 
178.3932 
176.9705 
176.1066 
174.2258 
size=19 
size=21 
size=31 
size=41 
-bandi- 
150.7933 
147.8115 
129.5668 
118.4827 
-band2- 
168.0672 
164.1161 
141.8316 
128.638 
-band3- 
193.1976 
188.7883 
163.6851 
148.5688 
-band4- 
122.0708 
122.2362 
115.5704 
109.892 
-band5- 
134.7398 
134.6939 
126.3652 
119.1299 
-band7- 
171.8355 
171.0026 
159.2047 
149.9347 
-bandl- mean the combination of band51-bandl-band54 
Table 1. OIF value statistical table 
9n Ina ¿no