Gain and Bias are two basic parameters for the reflectance calculation of TM images. The deviations of the two parameters will lead to the erroneous results of radiometric calibration, radiometric correction and atmospheric correction. Because of the ageing of Landsat-5 TM and the decrease of the radiation accuracy to some extent, the USGS has published different parameters of radiation correction in the past few years. Based on the Gain and Bias values acquired from the USGS and the image head file, the authors calculated the surface reflectance of water and vegetation and compared the calculation result with the standard reflectance. The results show that the Gain and Bias acquired from the head file is the best way. If the image has no head file, the Gain and Bias offered by the USGS could be used.