A new method to estimate composite top of the atmosphere (TOA) visible clear-sky reflectances for wide narrow band geostationary satellites such as the Meteosat constellation is presented. This method relies on some a priori knowledge of angular variations of TOA broadband reflectances associated with clear-sky conditions above mean surface types through the use of the clear-sky Cloud and the Earth's Radiant Energy System (CERES) shortwave broadband angular dependency models (ADMs). Each pixel (or Earth location) viewed from such geostationary imager at a given daytime is associated with a reflectance time series made up of its chronological daily measurements. This time series can be seen as a clear-sky visible narrow band reflectance curve of the associated pixel surface plus an additive random noise modeling cloudy conditions above it. On the basis of this assumption, TOA clear-sky broadband reflectances extracted from the CERES ADMs are used to compute curve-driven fifth percentiles on these time series in order to estimate the TOA clear-sky visible narrow band reflectance curves for all pixels, while the percentile approach exhibits only a reduced sensitivity to cloud shadows. Benefits of our method are discussed with respect to its application to 7 months of Meteosat-7 daytime visible narrow band measurements. Finally, the performance of our algorithm is assessed through comparisons with its predicted and associated International Cloud Climatology Project DX clear-sky values with respect to a visually generated clear-sky pixels database.