To handle the increasing penetration rate of intermittent energies such as photo-voltaic power plants, grid managers require accurate forecasts of the primary resources. In this paper, we propose a new method to forecast the Global Horizontal Irradiance (GHI) up to 1 hour using HelioClim-3 maps as only inputs. HelioClim-3 maps are Solar Surface Irradiance (SSI) maps derived from geostationary satellite images through the Heliosat-2 method. We use a block matching algorithm to retrieve the motion vector field, allowing us to anticipate the future on site GHI values. The results were then corrected following a statistical post-processing step tuned on a training data set. The model has been tested on a one year data set acquired on the SIRTA site, Polytechnique campus, France. Besides the very promising performances of the proposed method compared to the persistence model, we extracted additional forecast information from the SSI maps like the minimum and maximum expected values, which enlighten the model's potential and limitations.