(b) Cross-Correlation
5. Conclusions
G) The precisions of matching by cross-correlation
(a) Least Squares Method for images in which no distortions exist are only
marginally worse than for the least squares
(1) Tests on the precision of image matching by the method.
method of least squares indicate values as high
as .02 pixel can be achieved with cross and (ii) The effects of distortions between the two
circular features, but precisions will deteriorate images result in a rapid deterioration in the
for quantization levels less than 5 bits. quality of matching by the cross-correlation
method.
(11) The extent of image blur has only a marginal
effect on the precision of matching, with Ref
precisions improving slightly as the blur eierences
increases up to that described by a Gaussian
spread function with a 26 width of 501m. Beyer, H. (1992) 'High Precision Real-time
is 4 3 Photogrammetry’. PhD Dissertation Institute of
(iii) The overall differences in the quality of Geodesy and Photogrammetry ETH-Zurich.
matching for the cross and circle features are
only marginal, with generally better results Foerstner W. (1982) 'On Geometric Precision of Digital
being obtained for crosses than for circles. Correlation. Int. Arch. Photgram. & Rem Sens Vol.
E ; 24-3 pp 176-189.
(iv) Scale variations between the two images greater
than 10% will lead to a significant decrease in Trinder J.C. (1989) "Precision of Digital Target Location'.
the precision of image matching. Photogramm. Eng. and Rem. Sens Vol 55 pp 883-886.
(v) There is good agreement between the Trinder J.C., T. Tjugiarto & B.E. Donnelly (1990). 'A
previously published theoretical precisions and Digital Photogrammetry System for Close Range'. Aust.
those obtained by simulation in this study. J. Geod. Photogram. & Surv No. 53 pp 1-13.
823