Atmospheric correction is important for land remote sensing application. In data processing over multi-spectral remote sensing land image, an atmospheric correction method is presented. The method is based on 6S radiative transfer code and an empirical method of decreasing adjacency effect. To evaluate this correction procedure, FLAASH (another mature atmospheric correction model) is processed as a comparison. Results suggest that the quality of corrected image is improved apparently, the accuracy of the method and FLAASH remain about 10%.