pore
m sat-
eal
inges
| fea-
from
em,
y win-
'lassi-
n…al-
to the
P ob^
j im-
ispec-
1eces-
iture
3 also
nputa-
red
1d the
tail,
Por
ne to-
essen-
ay).
xel and
eeds a
es ad-
sa
struc-
depend
2. Basic Theoretical Concept
Contrary to the aim of the adjustment computation, it is not
intended here to determine the best fitting transformation
funetion between the two feature vectors x and y. A linear
funetional model is rather hypothetically assumed and the
residuals, resulting after the adjustment are examined as to
random or systematic deviations from this best fitting
straight line. Thus, a conclusion can be drawn to a linear
or non-linear dependency of two pixels each from both fea-
ture vectors x and y. The formulae used for this purpose is
indicated by Wolf [2].
In the following, it is con-
y sidered that the x; and y;
are of the same accuracy. Ad-
ditionally we assume not to
minimize the D sd but the
sum of the squared vertical
distances [vv] of the meas-
y
ured points from the best fit-
Fig.1 ting straight line G (see Fig.1).
As the adjusted straight line fits the central point S of
the measured points, one obtains the following observation
equations:
Vi = E; Sin 9 — N,°C0S 9 \ d
; = TE = iz)
with E1 = Xs eos and "n; = Yi —
Substituting
2 = (zn), b=022) = [nnl, ec= [5858] + {nn} and n
2
V(fzzi - [nniY? t BlEnlt,
one obtains for the ascent of the straight line
d
2