A method is presented for generating composite SAR imagery from a set of multiple radiometrically terrain-corrected (RTC) backscatter images and their co-registered local illuminated areas. The process is implemented in map geometry: the resolution of each image is estimated locally as the inverse of the local contributing area and used to weight the contributions from all available observations. Some potential contributors could be out-of-swath or occluded by radar shadow. The resulting composite backscatter image can trade off temporal resolution for improved local spatial resolution while simultaneously increasing the local number of looks. The composite image is characterized by more homogenous properties (resolution, noise) in comparison to single acquisitions. The technique is demonstrated using ScanSAR data from ENVISAT ASAR and Radarsat-2.