latle 7. influence of Model's beformation
C s odel Ch ck Foi nts | on ce RMS Z ————————-——
tese | e No Correction | Correction Done
| 10 1.87 m 0.89 m
ere ee ee NT TT
[450 11 1.62 m 0.84 m
|
- BX ra ERR TR OAL
: 4 12 2.14 m 0.84 m
2.2 Resampliing along the epipolar lines
bpipolar line is a well known conception to photogrammetrists.
Using this concept a inherent two- dimensional correlation task can
be reduced to one-dimensional matching problem. It is extremely
important to rearrange image data along epipolar lines, because
only under this circumstance it becomes possible to process a large
amount of image data by tne use of a small capability of memory.
In accordance with traditional way resampling along epipolar lines
is treated as a digital rectification problem, in which four pixels
have to be involved for extracting one resampled pixel (Fig 3.a).
lt is time-consuming for bilinear interpolation.
Epipolar line
277
T.
NIS
p
(a)
Fig. 5. Resampling along epipolar lines
In the SODAMS a new approach has been available, in which only two
neighbordhood pixels are needed to derive one pixel on epipolar
line (Fig.5.b). In fact a further step can be taken to simplify the
resampling procedure with tne help of the nearest neighborhood
method instead of linear interpolation (Fig.3.c). With this approxi-
mate measure there is no effect on further processing -- correlation.
Therefor, the resampling along epipolar lines can be simplified as
a procedure to pick up corresponding groups of pixels from origi-
nally digitized images (Fig.3.c). It is powerful and timesaving to
run the program written in such way mentioned above.
2.3 Digital correlation
- 676 -