Radargrammetric improvements: A multi-window approach

This paper deals with the relevance of using stereoscopic radar images in order to retrieve the relief of terrain. Firstly, the basic characteristics of the radargrammetry are described. Thus, we present the results of the radargrammetric processing using the ZNCC procedure and applied to radar images. These images are recorded by the SIR-C mission over the French Alps. The results show that the image matching can fail, especially in foreshortened areas. So, we expose two different improvement methods using several correlation windows in order to cancel the reconstruction errors. The first improvement take advantage of a multi-window approach to combine information multiplying the correlation surfaces obtained for each correlation window size. The second improvement is based on another multi-window approach that makes it possible to get correlation windows adapted to the foreshortened areas. Finally, we combine these two improvement methods to show that it´s possible to make the disparity map more reliable for the first step of the pyramidal scheme.