relief-corrected affine transformation (for the rationale
for using this model and other details, see Baltsavias et
al., 2001). Firstly, the used GCP object coordinates are
reduced to a reference height level (the height selection
is irrelevant, but here a height of 0 m was used). Then,
an affine transformation between these reduced GCPs
and their pixel coordinates is estimated by least
squares. A minimum of three GCPs is required,
whereby their spatial distribution can be relaxed, and
extrapolation does not lead to errors. This model has
been used with several IKONOS Geo images for
orthorectification (Baltsavias et al., 2001). The second
model compared to the first one has several advantages:
computation time is lower, less GCPs with not-sostringent
spatial distribution criteria are needed, and
extrapolation errors do not occur. Thus, it was even
anticipated that the second model would lead to higher
orthophoto accuracy.
The first model was estimated once with all 38
GCPs and a second time with 28 GCPs, while the