Different from visible signals, thermal infrared radiances depend on both temperature and emissivity. It is a key problem for us to separate temperature and emissivity in thermal infrared remote sensing re- search. Another difficulty encountered in the retrieval of surface temperature is the correction of downwelling sky irradiance, because it is closely related to surface emissivity. When emissivity is un- known, the downwelling sky irradiance is difficult to be removed. In this paper, we introduce a correc- tion term of downwelling sky irradiance developed by Li and Becker into Wien’s approximation, to de- rive an improved ALPHA difference spectrum which is independent of temperature, and furthermore develop a correction term to remove the error of Wien’s approximation. Under the support of the above work, attractive features of Alpha derived emissivity method and ASTER TES algorithm are combined together to acquire a new Improved TES algorithm based on Corrected ALPHA Difference Spectrum (ICADS TES). Because a multi-band inversion technique is applied, and the operations of band ratios and differences are included in the algorithm, it can partly remove the influence of atmosphere and noises. Numerical simulation experiments show that for various combinations of atmosphere, land covers and surface temperatures, the algorithm is applicable and stable. Its accuracy for temperature is 0―1.5 K, and that for emissivity is 0―0.015. Compared with current TES algorithms, our method has clear physical meaning, is easy to be implemented, and is applicable for a wide temperature range and surface types. The results are not influenced by the directional characteristic of emissivity. Because ICADS TES does not need the support of a priori information of surface types, it is also not influenced by the accuracy of classification and the problem of mixture pixels. Compared with our former TES algorithm based on corrected Alpha difference spectra (CADS TES), the new algorithm takes the effect of downwelling atmospheric radiation into account. When the quantity of atmosphere radiation can be estimated precisely, the performance of ICADS TES is much better. Different from visible signals, thermal infrared radiations depend on both temperature and emissivity. It is a key problem for us to separate temperature and emissivity in thermal infrared remote sensing re- search. Another difficulty encountered in the retrieval of surface temperature is the correction of downwelling sky irradiance, because it is closely related to surface emissivity. When emissivity is un- known, the downwelling sky irradiance is difficult to be removed. In this paper, we introduce a correc- tion term of downwelling sky irradiance developed by Li and Becker into Wien’s approximation, to de rive an improved ALPHA difference spectrum which is independent of temperature, and develop a correction term to remove the error of Wien’s approximation. Under the support of the above work, attractive features of Alpha derived emissivity method and ASTER TES algorithm are combined together to acquire a new Improved TES algorithm based on Corrected ALPHA Difference Spectrum (ICADS TES). Because a multi-band inversion technique is applied, and the operations of band ratios and differences are included in the algorithm, it can partly remove the influence of atmosphere and noises. Numerical simulation experiments show that for various combinations of Its accuracy for temperature is 0-1.5 K, and that for emissivity is 0-0.015. Compared with current TES algorithms, our method has clear physical meaning, is easy to be implemented, and is applicable for a wide temperature range and surface types. The results are not influenced by the directional characteristic of emissivity. Because ICADS TES does not need the support of a priori information of surface types, it is also not influenced by the Accuracy of classification and the problem of mixture pixels. Compared with our former TES algorithm based on corrected Alpha difference spectra (CADS TES), the new algori thm takes the effect of downwelling atmospheric radiation into account. When the quantity of atmosphere radiation can be estimated precisely, the performance of ICADS TES is much better.