293 
4.2 Result analysis 
From the processing result above, it could be seen that the 
segmentation algorithm proposed in this paper is feasible. 
However, the results also have some problems. There are many 
small scraps in some region of segmentation, which should be a 
whole cluster region. It may be caused by the following points 
that are still in improvement. 
The first problem is to choose suitable value for displacement d. 
Especially for different kind of images, How to give a suitable 
value for displacement d is still a difficult thing. 
The second problem is to calculate the initial value for 
clustering. The initial value of clustering is vital to the 
segmentation. However, it is difficult to find a common method 
to give initial value for different images. 
The third problem is to optimize the weights between texture 
features. For example, the mean of energy is 0.000137 while 
the mean of contrast is 1766.045. Therefore, it is necessary to 
produce suitable weights for texture features automatically. 
However, due to the difference of the images to be processed, it 
is hard to find an effective method. 
red by 
ge. The 
is the 
5 CONCLUSIONS 
By use of texture features, this algorithm proposed in this paper 
can overcome the shortcoming of the traditional segmentation 
methods. From the segmentation results of different original 
images, it can be seen that texture features are useful in high 
spatial resolution RS image processing. 
REFERENCES 
LIU L.F., 2003. Texture Analysis Methods Used In Remote 
Sensing Images. Remote sensing technology and application, 
18(6):441-447 
Ming D.P., 2004. Remote sensing image segmentation 
algorithm based on simplified random field model. Computer 
engineering and application, (26): 28-30 
HUANG X., 2006. Methods for Classification of the High 
Spatial Resolution Remotely Sensed Images Based on Wavelet 
Transform. Geomatics and Information Science of Wunan 
University, 31(12): 1055-1058 
Su J.Y., 2004. Semi-automatic Extraction Technique of 
Residential Area in High Resolution Remote Sensing Image. 
Geomatics and Information Science of Wunan University, 
29(9):791-795 
Zhou F., 2001. Studies on texture analysis, Ph.D dissertation of 
Peking University, China.