234 
PHOTOGRAMMETRIC ENGINEERING 
Fig. 10. DAMC Sequential Program System flow chart. 
magnetic tapes. One tape contains the packed 
binary-gray shades from the two upper 
quadrants of the stereo pair, and the other, 
those from the lower half model. The first 
phase of the rectification program computes 
the x and y data shift lists required to carry 
out rectification and scaling. It also accepts 
data on the stereo images of the left and right 
nadir-points and produces information needed 
for the ortho-correction program. 
The scanned data and rectification lists are 
used to modify the scanned data in both the x 
and y directions. An output area is reserved in 
core memory to store a maximum of 100 
scans. Once this area filled with rectified data, 
the maximum number of full scans are written 
on an output tape. The incomplete section is 
moved down to become the start of the next 
rectified block. This process continues until 
the entire input data tape is exhausted and 
the program is complete and ready for a cor 
relation program to measure x-parallax 
differences. 
Successive Parallax Correlation Program 
The correlation method of successive 
parallaxapproximations is based on twofunda- 
mental ideas: convergence to a solution by 
stages of increasingly accurate parallax 
measurements, and use of statistical correla 
tion of image data to find matching areas. The 
idea of convergence is implemented by using 
digital images of reduced resolution. The re 
duced images can be computed quickly. A 
single area of one image is matched by com 
puting correlation coefficients for several 
overlapping areas of the other photo. The 
greatest correlation coefficient is the criterion 
of best match. A correlation coefficient be 
tween x and y is conventionally represented 
by the letter “r" as defined by 
co-variance (x, y) 
r = 
S x -S u 
x = Z)(density level of left photo point) 
y = //(density level of right photo) 
The denominator of the formula is the prod 
uct of the sample deviations for x and y, the 
density levels in the sample areas of left and 
right photos, respectively. The formula used 
for the sample deviation is: 
Sr = Vp - x 2 , S u = ^y 2 - ? 
where 
1 ^ , _ 1 » 
x = — and x 2 = — : 2^ -T, 2 
n l n 1