The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B3b. Beijing 2008 
472 
6. CONCLUSIONS 
In this paper, the application of the wavelet analysis method on 
extracting the building from stereo images is described. Some 
new ideas and methods about how to solve several key 
questions are put forward. The results are summarized as 
on compact support, good local regularity and smoothness. 
Furthermore they are provided with very good multi-scales 
properties. When the scale is small, there are plentiful details in 
the detected result image. With large scales some details in the 
image are neglected and the main edge features are intensified. 
So we can realize optimal edge detection of buildings. 
2) The algorithm of the lifting scheme ASWlet wavelet 
decomposition and reconstruction of image is studied, and the 
prediction function and update function were designed. The 
method could maintain the on-site computation property of 
lifted wavelets. In the meantime, it has strong expressive ability 
as for the high frequency feature of the three constituents on the 
decomposed image (vertical, horizontal and diagonal direction). 
The result of wavelet transform could be very useful to the next 
image match. 
3) The basic principle and measure of semi-automatic 
building extraction is discussed and a new method of automatic 
comer point recognition is proposed. The method depends on 
binary image morphology after feature extraction, instead of 
recognizing the comer points on gray-scale image. So it can 
locate the comers accurately without gray and geometry 
threshold. Meanwhile, the description of invariant feature 
parameters and the feature matching based wavelet transform is 
adopted, which provides a fundamental basis for the 
semi-automatic extraction of the building from the remote 
sensing image. 
4) On the basis of multi-scale edge detection based wavelet 
analysis we have studied comer point recognition and feature 
matching, the method of building model automatic extraction 
and 3D reconstruction on stereo images. 
Based on remote-sensing image, we have investigated as well 
wavelet applications in different critical problems for the 
building automatic extraction. Experiments have proved the 
theories that we proposed were feasible and practical with high 
accuracy. The methods developed here are shown to be effective 
follows: 
1) The method of multi-scale feature extraction based on 
wavelet analysis was studied, and a new anti-symmetrical 
wavelet (ASWlet) function and corresponding filters have been 
proposed, which are symmetry, approximati 
for improving efficiency on 3D building automatic acquisition 
and for speeding up the progress of digital city construction. 
REFERENCE 
Cheng, W. 2002. Wavelet Analysis and Application on 
Image Processing. Science Press, Beijing. 
Chen, Y. 2000. Wavelet- Based Image Matching for 
Different Sensors. Inter. Archives of ISPRS, XXXIII B3/1, 
pp. 177-182. 
Chen Y. 2003. Digital Photogrammetry for Remote Sensing 
Images. Tongji University Press, Shanghai, pp.204 -206. 
Chen, Y., Y. Lin 2006. Image matching and translation base 
on lifting scheme wavelet. Acta Geodaetica et Catographica 
Sinica, (1), pp: 19-23. 
Cheng, Z. 1998. Wavelet Analysis Algorithms and 
Application. Xi’An Jiaotong University Press, Xi’An. 
Lin, Y., Y. Chen 2004. The construction of a Wavelet filter 
and multiscale edge detection. Computer Engineering and 
Applications, 40(35), pp.22-40. 
Lin, Y., Y. Chen 2007. A direction profile detection for edge 
extraction with a Linear Lifting Scheme Wavelet. Journal of 
Tongji University(Natural Science), 35(12) 
Marr, D. 1982. Vision. H. Freeman and Co. 
Peng, Y. 1999. Wavelet transform and Engineering 
Application. Science Press, Beijing. 
Smith,S.M., Brady,J.M. 1997. New Approach to Low Level 
Image Processing. Int. Journal of Computer vision, 23(1).