An optimal interpolation scheme for producing a DEM from the automated stereo-matching of full-scale SPOT images

Eight interpolation methods are implemented and tested over a 60 km × 60 km region in Korea, using the results from automatic stereo-matching of SPOT images. First, with the simulated elevation data from digitised contours of paper maps, Kriging and Multiquadric methods show higher performance: with the accuracy better than 3 m than other methods and the speed of ~400 s on a Sparc 20 machine. Second, in a realistic case when the elevation derived from SPOT images is used, Kriging and Gaussian perform better than other methods by 1 to 5 m in terms of accuracy. Due to the errors in the SPOT stereo-match output, Gaussian appears to perform better than Kriging. As a third experiment, we consider large regions of matchfailure due to cloud cover, image skewness, and grey-value differences. To prevent interpolation on these regions, a ‘centre-of-gravity’ within each search region of interpolation is used as a rejection criterion.