Atmospheric aerosol is an important factor of the Earth's radiation budget. The aerosol optical depth (AOD) is also the key parameter in generating surface products from remotely sensed data. An image-based retrieval algorithm of aerosol characteristics and surface reflectance is used to retrieve the AOD from SPOT satellite images in this paper. The accuracy of retrieved AOD is assessed using sunphotometer measurements. SPOT satellite images in Jhongli, Taoyuan county are used to testify the algorithm. The results show that the root-mean-square error (RMSE) of the retrieved AOD at 0.55 mu m is 0.10. Over the range of measured AOD 0.08 similar to 0.34, the mean relative error is 49%. The RMSE of the retrieval is very sensitive to assumed DT reflectance: it can be reduced to 0.067, when assumed DT reflectance in green band is set 0.035, instead of 0.03. Urban aerosol model is not suitable in this test area, because of its high absorption. The RMSE of retrieved AOD is insensitive to continental, maritime and biomass-burning aerosol models, since the deviation of RMSE of the retrieved AOD using these three models is within 0.02. The errors are also shown to be independent of the measured AODs. More observations over different locations and canopy species are required to testify the algorithm in the future.