The maximum accuracy potential should be even higher if one considers that the reference DEM was about as accurate as the one derived from IKONOS, that a significant bias (mean with sign), which could be potentially removed, existed, and that the RMS and mean with sign were influenced by large blunders, caused by faint remaining clouds which were not manually excluded, and thus led to much higher elevations in the matching results. More details on these investigations can be found in Zhang et al. (2002), while an example of the DSM generated by matching is shown in Fig. 7.