New method for backscatter model estimation and elevation map computation using radarclinometry

Synthetic aperture radar (SAR) imagery is now well-known as it is actively used in various remote sensing domains such as oceanography, geology, environment surveillance, cartography, etc. In this paper, we describe a new method for computing an elevation map from a single SAR image. Our method begins by reconstructing an approximate elevation map of the imaged ground using radarclinometry. This reconstruction relies on a Lambertian assumption for the backscatter model, which, as our experiments have shown, is not particularly accurate. Therefore, once we have computed the initial estimation, we proceed to estimate a more accurate backscatter model for the imaged ground. After a brief description of the relevant reflection mechanisms, we show that the theoretical considerations are too complex to be completely and accurately modeled. We therefore present two methods for performing the backscatter estimation. The first method is based on empirical results. The second method makes use of the approximate elevation map obtained by radarclinometry. In the latter case, we subsequently compute an improved elevation map using the more accurate backscatter model.