Ying Chen 
  
mean square error of matching was less than 0.5 pixel. 
       
SPOT image CCD image SPOT image CCD image SPOT image CCD image 
a feature without filter b feature with filter c matching result 
Fig3 Feature extraction and matching result 
Experiment results demonstrate: 
a. The applications of Wavelet transform for matching between different sensor images have brought about the 
good effect. Specially, using Lipschitz exponents is propitious to reduce noise during edge Detection. 
b. Edge direction profile detection method makes two dimension problems to one dimension processing. It is 
simple and efficient. 
c. Line moment is independent to scale and rotation. Initial! matching using moment parameters and available 
rigorous constraints could provide enough accurate initial value for least square image matching. Finally exact 
sub-pixel results would be attainable by least square matching. 
REFERENCES 
Mallat S, Hwang W L. Singularity Detection and Processing with Wavelet. TEEE Trans. On Information Theory 
Vol.38,No.2,March 1992,617-643. 
Jinfei Wang: LIND-A System for Automated Linear Feature Detection and Analysis. CANADIAN Journal of 
Remote Sensing ,1993. 
Lide Wu: Computer Vision, The Publish House of FuDan University, China,1993. 
Huili, B.S.Manjunath, S.K.Mitra: 4 Contour-Based Approach to Multisensor Image Registration. 1EEE 
Transactions on Image Processing, Vol.4,N0.3,3.1995. 
W.Wen and A. Lozzi: Recognition and Inspection of Manufactured Parts Using Line Moment of Their Rounders, 
PR. Vol.26,No.10,1993,pp1461-1471. 
Ying Chen, A Real Time Image Matching Based on Wavelet. Journal of Image and Graphics , 
1998.12,pp1011-1014 | 
Ying Chen, A feature-based and least square Image matching with Costraints. Journal of Image and graphics 
1998.4, pp299-303 
  
182 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.