First, a brief introductory review of previous works on atmospheric correction of satellite imagery over lands was given. Then we presented the theoretical basis of the atmospheric correction algorithm which removes both atmospheric scattering and adjacent effects from the satellite image data with a fine ground resolution. We proposed a simple algorithm for estimating aerosol optical thickness from satellite image data by using dark target pixels and applied it to Landsat TM and ADEOS AVNIR data. As for the algorithm validation, several simultaneous ground and sky measurements data were used. We found that the satisfactory results were obtained in the cases of sub-scene imagery, whereas the agreement was not so good in the full-scene case due to local aerosol's spatial variation.