Topography introduces radiometric biases which impact the classification process. Terrain variation affect both the position of a given point and the brightness of the radar return. This can greatly bias interpretation of the scene and classification results. Small in [R-7] proposed a terrain flattening method that removes the radiometric variability associated with topography while leaving the radiometric variability associated with land cover.In the conventional expression of radar backscattering coefficients, no Earth mode is considered. Therefore, although the position, or geometry of the backscatter estimate has been corrected in GTC products, the radiometry of the resulting image remains ellipsoid-model based. In the terrain flattening method proposed by Small [R-7], a radiometric image is simulated using orbital information and the digital elevation model (DEM). In the simulated radiometric image, each pixel represents the local illuminated area for the point in the image. Finally, the terrain flattened gamma naught scatter is produced through a normalization scheme using the simulated radiometric image.In the calculation of local illuminated area for a given point, a DEM is used. E00 is the given point on the ground, E01, E10, and E11 are the neighbouring points to the east and north. With elevations from the DEM, we have the true ground area defined by points T00, T01, T10 and T11. To conform to the definition of γ0 backscatter, the area of the terrain facet triangles should first be projected onto the plane perpendicular to the local slant range direction. Then the local area is the sum of the areas of two triangles, i.e. P1(P00-P01-P10) and P2(P11-P01-P10). Define semi-perimeters of the triangles P1 and P2 as the follows: then the corresponding illuminated area can be calculated using Heron's formula. Terrain flattening is a radiometric correction of the synthetic aperture radar image. It normalizes the radar backscatter for the terrain variation with a radiometric image simulated using the accurate radar model for SAR acquisition and the Digital Elevation Model (DEM) of the target area. The major processing steps are as follows: 1. Calibrate the raw SAR image to beta naught; 2. Simulate an image of the illuminated area using SAR acquisition model and DEM; 3. Normalize the radar backscatter in beta naught with the simulated image to produce the radiometrically terrain corrected gamma naught image. The simulation of the illuminated area is the key processing step. The illuminated area is computed in the map domain for each target cell using SAR acquisition geometry. The computed illumination area is then distributed from the map geometry into pixels in the radar geometry.Distribute computed illumination area from map domain into pixels in the image domainFor better useability, it can be a simple modification to first perform the calibration to Beta0 within the Terrain Flattening operator if it has not already been done.Furthermore, the terrain flattened image is not available in the map domain. Currently users need to perform terrain flattening and terrain correction consecutively in order to obtain the map domain terrain flattened images. This can lead to performance limitation within a single graph as well as large memory consumption.In the terrain flattening algorithm proposed in [R-7], the simulated image of the local illuminated area is computed originally in the map domain. The map domain simulated image is later mapped into the image domain to get the image domain simulated image. This is because the final terrain flattening is performed in the image domain. If the terrain flattening is performed in the map domain, then the mapping of the simulated image from map domain to image domain is no longer needed. Instead, we only need to perform terrain correction to generate the image in map domain. By combining the terrain flattening and terrain correction, users will need to run one operator, instead of two, to get the terrain flattened images in map domain. A lot of redundant calculations can be avoided and memory consumption can be reduced.