Herbert Jahn
Figure 1. WAAC stereo pair “Village”
The bright horizontal lines show the places where a cross section of both images were taken. Figure 2 shows the gray
value profiles along those lines. One can see the varying disparity. Figure 3 shows the profiles after matching. A good
coincidence of both profiles now is observed.
À ] E n
400 A "n i al 3 sok | | | m . E
3 Hi M à i h bef : : E M | n f Sas » ?
© "CT J 1 + 9 F Î qd
> 300 5B jw! h a à = > 300 ad | : 3
PO bey NE ys Moog i 1 — E key | i lys! Me Lar
hp * 3 E 3
200 uu Wa i 200 - N : 3
or = ; zu n m i e t es = Ko : To er
Figure 2. Gray value profiles Figure 3. Gray value profiles after matching
The disparity s, along the horizontal line is displayed in figure 4. This seems to be a useful result although ground truth
is not available. Figure 5 shows the s,- image.
A 8 T TUTTI T T T
M
} + fl | ]
: kb 1
di Ti
1 : [ M | | | ^ q
| z | | | LA A 2 | ~ m
i 8 2. | | \ iM | | \ 4l n a
I $ [ | \ | M b] \n | \ Í |
9 e | /\ |
| Li / | V / V Los]
| 0 MI Ho | i M \ W zd
| - ||
[ L Un
| -2F | | x
-4 L.. A 1 L d L L L
0 50 100 150 200 250 300
1 column i
; Figure 4. Disparity profile Figure 5. Disparity image
The stereo pair of figure 1 is not strictly epipolar. After matching in x- direction as was shown here there remain small
shifts in y- direction in some image regions caused by aircraft attitude disturbances. These shifts can be nearly removed
by applying the algorithm in y- direction. A red-green coded overlay of both images reveals this but that is not shown
here because of limitations of printing space. Some areas with matching errors remain especially near buildings where
occlusions occur. Furthermore, the disparities seem to be too smooth.
Ez
442 International Archives of Photogrammetry and Remote Sensing. Vol. XXXIII, Part B3. Amsterdam 2000.