There are two methods: 
a) Comparison of 2 single columns next to one another 
S1 = 2 a(x1,y) $82 - 2. g(x2, y) , where x2 
1 1 i 
XE +3 =, 
duc. MY 
[51-52 |] 
b) Division of the whole window into 2 equal parts (Yakimovsky's 
method, see also /7/) 
1. Comparison of simple summations, i.e. brightness difference 
between the subsections 
S1 - 2. 2.9(x1,y) x S -2: 2.302. y) , where x2 = x1 + MX/2 
1: Did 
i ms. MY 
j 284.4 MX? 
2. Comparison using variances, i.e. structural difference between 
the subsections 
Mean grey values : 9,7 2 S1 / (MX NO ; 957 2 S2 / (MX MY) 
Variances: V1 2424 s (x1, y) -$,)* siae 2.2. s (x2, y) hb 
IV» ]s1-s2]| resp. | v1 - v2 | , depending on the local 
  
  
variations 
Results 0 : Regular grid 
la, 1b : Interest operators as above 
Unknowns : AO = parallax HO = intensity shift 
Al = tilt H1 = intensity scale 
A2 = shear 
Operator : 0 1a 1b 
Least squares : SD SD R SD R 
AO (0.01 pixel) : 11.5 2.2» «0.61 7.80: =0.61 
A1 (0.01) 24 1.7 00.25 1.6 — -0.16 
A2 ( " ) : 1.9 1.2. 5: ~0.66 1.1 -0.47 
HO (grey value) : 2.0 1.1 0.8 
Hi (0.01) : 6.4 4.3 2.0 
nope a: oft a! de I aR 
Cross correlation : CC CC R CC R 
mean maximum ; 0.72 0.91 +0.44 0.95 . 40.51 
-568 -