Abstract: The pixel dichotomy model based on normalized difference vegetation index (NDVI) was adopted to extract the vegetation coverage from original DN values of TM image, top of atmospheric, surface reflectance with MORTRAN radioactive transfer model and dark-object method of simple transmission model, which were compared with observed values of vegetation coverage, in order to analyze the effects of atmospheric correction on remote sensing estimation of vegetation coverage. The target area was Jiulianshan reserve located in southern Jiangxi province, an eastern province of China. Result showed that although the correction principle of two atmospheric correction approaches, including FLAASH and DOS, was different, but they performed significant for eliminating atmospheric effects, and the accuracy of remote sensing estimation on vegetation coverage was effectively improved, based on the degree of fitting to observed values in the sample plots. The vegetation coverage estimated by top of atmospheric after radiation calibration and original image DN values were not significantly different, illustrating that the improvement of vegetation coverage estimation accuracy was derived from the atmospheric correction, and the effect of radiation calibration on vegetation coverage estimation was very small. Therefore, atmospheric correction is meaningful for improving estimation accuracy in subtropical areas with high vegetation coverage.