local variances in the pattern and search matrix, 
c) the correlation intensity coefficient (like b) with a weighting parameter P 
depending on the global variances in pattern and search matrix, 2 
d) the Laplace coefficient, i.e. the sum of the absolute differences of the 
image signals and 
e) the phase correlation coefficient, i.e. the inverse Fourier transform of the 
phase cross spectrum of the image signals; this function is derived for mis- 
sile guidance in military application (Kuglin et al. 1979). 
Functions a), b), c) and e) have to be maximized, function d) has to be mini- 
mized to compute the most probable location of the pattern in the search matrix. 
Function e) reqüires a twodimensional Fast-Fourier-Transform (FFT), that has 
been installed in the MOBI-software-package and allows the FFT of images with 
arbitrary size (up to 2048 x 2048 picture elements). 
IMAGE MATERIAL 
  
As reference image a digitized section of a RMK (RMK - Reihenmegkammer) aerial 
photograph with an original scale of 1 : 50 000 has been chosen (fig. 1) 
  
200 - 
   
Fig. 1: Aerial reference image Fig. 2: Image with SNR - 2.6 
  
It shows a part of Hannover. This reference image is assumed to be noise-free, 100 . 
it contains the undistorted image signal. Now random noise in different magni- 
tudes (Gaussian and uniform distributed) is added to the reference signal. The 
SNR of the noisy images lies between 60 and 0.8. Fig. 2 and 3 show two noisy 
images with an SNR = 2.6 and 1.0 respectively. 
The differences can be clearly seen in fig. 4 - 6 that show the same line in the 
reference and the 2 distorted images. 
In the reference image 20 correlation points for the accuracy test has been 
chosen. The correlation points are the centers of 11 x 11 pattern matrices. In 
the noisy images the search matrices have the same centers and a size of e 
30 x 30 pixel. Fig. 7 shows the reference image with the marked correlation ; 
points. The white crosses indicate high variances in the pattern matrix (mean Fig. 
value c" = 198.4), the black crosses small variances (52 = 23.6). 
72 
PY TÉ 
bip