An image-based retrieval algorithm of aerosol characteristics and surface reflectance for satellite images is proposed. By assuming the Junge size distributed aerosol in the atmosphere and feeding back the new Junge parameter, not only the aerosol optical depth but also the Junge parameter, single-scattering albedo and phase function can be iteratively derived and converged from digital counts of dense dark vegetation in the green and red bands of SPOT satellite image. These retrieved aerosol characteristics are considered to be more physically related to the image itself than estimated. To prevent over-estimating aerosol optical depth, multiple scattering of path radiance for aerosol is taken into consideration. Although in lack of field measurement, the algorithm is evaluated and proved to be useful by simulation. Its sensitivities to assumed dense dark vegetation reflectances, log-normal size distribution, initial assumed power of Junge size distribution, refractive index and radius range are also studied. The retrieved aerosol optical characteristics are further used to derive surface reflectance of SPOT satellite image by a proposed atmospheric correction model for non-uniform and Lambertian surface. Without atmospheric correction, the RMSE between the apparent and field-measured reflectance is 0·076 for reflectance range 0·0 to 0·6. The RMSE between the derived and field-measured reflectance is greatly reduced to 0·022, if the image is atmospheric-corrected.