Abstract
The authors evaluated the actual performance of some kinds of regression models for correcting image distortion due to surface topography in various kinds of satellite images including SAR data. For optical sensor images like SPOT/HRV, it was verified that a simple regression model in which only one height term is added to affine or pseudo-affine transformation can be applied with the error of about one pixel. On the other hand SAR data by RADARSAT resulted in the error twice of that for optical data, which was considered due to a difficulty in tie point selection between SAR and optical sensor images. Next, this technique was extended to height measurement. A case study with a SPOT panchromatic stereo pair verified that the similar regression model as that for height correction is possible to be applied for height measurement and the partial revision of a conventional DEM was attempted by using this regressive method.
