inbul 2004 
sition were 
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B3. Istanbul 2004 
Table 1. Results using least-squares matching 
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
  
Coordinates Coordinates Affine Discrepancies Projective Discrepancies 
(master) (slave) ransformation (pixels) transformation (pixels) 
1 (381, 220) (379, 238) (379.6, 238.4) 0.69 (379.2, 238.0) 0.24 
2 (205, 350) (205, 369) (206.7, 367.7) 2.10 (208.3, 367.6) 3.62 
3 (91,285) (91, 304) (94.9, 303.8) 3.88 (93.7. 303.8) 2.73 
4 (112, 1073 (111, 125) (110 2,123 1) 2.07 (109.7, 123.4) 2.04 
5 (246, 39) (237,55) (237.1. 35.6) 0.65 (237.1, 56.6) 1.61 
Average 1.88 : 2.05 
Table 2. Results using cross correlation matching 
Coordinates Coordinates Affine Discrepancies Projective Discrepancies 
(master) (slave) ransformation (pixels) transformation (pixels) 
I (381, 220) (379, 238) (370.1. 233.3) 0.33 (370.0. 238.2) 0.22 
2 (205, 350) (205, 369) (205.3, 368.7) 0.48 (205.0, 369.0) 0.00 
3 (91, 285) (91, 304) (91.0, 304.0) 0.00 (91.0, 303.8) 0.24 
4 (112. 107) (111,125) (110.5, 125.1) 0.54 (110.4, 125.2) 0.64 
S (246, 39) (237, 55) (237.5, 54.9) 0.51 (237.2. 53.9) 1.10 
Average 0.38 0.44 
  
  
  
  
  
The experimental results are summarized in Tables | and 2 for 
the example shown in Figure 4. 
The results obtained using the cross correlation matching, all of 
which accomplished sub pixel accuracy, are better than those 
achieved using least-squares matching. It was initially assumed 
that projective function would result in the best improvement, 
however affine transformation was more effective in some cases. 
S. CONCLUSIONS 
The conclusions of this paper are as follows: 
e An improvement to the standard method of least-squares 
matching is possible, in regard to optimization. 
* The relationship between least-squares matching and 
cross correlation matching has been confirmed. 
* Cross correlation matching can be formulated to include 
image deformations. 
e An experimental comparison between least-squares 
matching and cross correlation matching by using 
IKONOS stereo imagery has been demonstrated. 
Geometrically constrained matching could be expected to 
further improve accuracy, though this paper has not dealt with 
geometric constraints. ^ Nevertheless, the cross correlation 
matching yielded positive results. It indicated that the method 
has potential to be a powerful matching strategy. 
Verification of the stability of the method needs to be further 
investigated by applying various conditions (illumination and 
location, including central city with high buildings, suburbs, 
mountain area, etc.). So far, this application has been relatively 
restricted. 
As mentioned above, our final goal is to make a contribution to 
the matching of multi-temporal and multi-resolution images, 
including IKONOS, QuickBird and aerial imagery. It will 
require a combination with other techniques, especially those 
related to change detection and resampling. 
605 
6. REFERENCES 
Ando, S., 2000. Consistent gradient operators. [EEE 
Transactions on Pattern Analysis and Machine Intelligence, 
22(3): 252-265. 
Brown, L. G., 1992. A survey of image registration techniques. 
ACM Computing Surveys, 24(4): 325-376. 
Fraser, C.S., Hanley, H.B. and Yamakawa, T., 2002. 3D 
geopositioning accuracy of IKONOS imagery. Photogramm. 
Record, 17(99): 465-479. 
Fraser, C.S. and Hanley, H.B., 2003. Bias compensation in 
rational functions for IKONOS satellite imagery. Photogramm. 
Engineering and Remote Sensing, 69(1): 53-57. 
Greene, W. H., 2000. Econometric Analysis. Prentice Hall, 
New Jersey, pp. 255-259. 
Gruen, A., 1985. Adaptive least squares correlation: a powerful 
image matching technique. South African Journal of 
Photogramm., Remote Sensing & Cartography, 14(3): 175-187. 
Gruen, A. and Baltsavias, E.P., 1988. Geometrically 
constrained multiphoto matching. Photogrammetric Eng. and 
Remote Sensing, 54(5): 633-641. 
Dare, P., Pendlbury, N., and Fraser, C.S., 2002. Digital 
orthomosaics as a source of control for geometrically correcting 
high resolution satellite imagery. Proceedings of 23rd Asian 
Conference on Remote Sensing, Kathmandu, Nepal, CD-ROM. 
Press, W. H., Teukolsky, S. A., Vetterling, W. T. and Flannery, 
B. P., 1988. Numerical Recipes in C. Cambridge University 
Press, Cambridge, pp. 681-688. 
Szeliski, R., 1996. Video mosaics for virtual environments. 
IEEE Compter Graphics and Applications, 16(2): 22-30.