MODTRAN model was used for the atmospheric correction of one HJ-1B/CCD2 image, and the effect of atmospheric correction was evaluated from the changes of spectral characteristics of typical ground objects,the comparison with the MODIS surface refluctance product,and the effect on normalized differential vegetation index ( NDVI). The results show that atmospheric correction eliminated the increase effect in visible bands and the absorption in near-infrared band. Atmospheric correction results and the MODIS surface reflectance product with high accuracy were highly consistent in the reflectance of vegetation, water and residents, and the average error of vegetation was 12.8%.According to the comparison of changing characteristics of NDVI before and after atmospheric correction, it could be found that atmospheric correction had corrected NDVI of mixed pixels and made it more reasonable. NDVI of each kind of ground objects improved, among which NDVI of vegetation increased most greatly,which can help differentiate vegetation from other ground objects. In a word, MODTRAN model has a good effect on atmospheric correction of HJ/CCD images.