Land cover classification requires both temporal and spatial information. Indeed, vegetation temporal evolution is necessary to discriminate the different land cover types. This information can be derived from coarse resolution sensors such as MERIS (300 × 300 m² pixel size), or SPOT/VGT (1 km² pixel size), whereas high resolution images, such as SPOT4/HRV ones (20×20 m² pixel size), contain the required spatial information. In this paper, a new method is proposed to perform an efficient land cover classification using these two kinds of remote sensing data. This method is based on Bayesian theory and on the linear mixture model permitting, through a simulated annealing algorithm, to perform a high resolution classification from a coarse resolution time series.