The cotton field infected by Verticillum wilt was investigated with both the multi-temporal TM images and the field survey simultaneously. A model of evaluating disease severity of the cotton was established by analyzing the correlation between spectral indices of TM image and severity level (SL) of the disease. The results indicated that with an increase of disease SLs, the values of spectral indices B2，B4，SATVI，OSAVI，MSAVI，TSAVI，SVNSWI，SNSWIa，SNSWIb，SVNI，DNSIa，DNSIb，NDSWIa，NDSWIb，RNSWIa，RNSWIb，DVNI，EVI，TVI，SAVI，DVI，NDVI，RVI and PVI declined slowly, B1，B3，B7 and RI increased gradually，NDGI increased at first and then decreased，while B5 exhibited a trend of decrease-increase with an increase of disease SLs. The SLs of disease were highly significantly positive correlated with the spectral indices values of B1，B3 and RI, highly significantly negative correlated with the spectral indice values of B4，OSAVI，MSAVI，TSAVI，SVNSWI，SNSWIa，SNSWIb，SVNI，DNSIa，DNSIb，NDSWIa，NDSWIb，RNSWIa，RNSWIb，DVNI，EVI，TVI，NDGI，SAVI，DVI，NDVI，RVI and PVI, significantly negative correlated with the spectral indice values of SATVI, and no significantly correlated with the spectral indice values of B2，B5 and B7. All of the eight spectral indices of TM image selected achieved significant correlation level. However, the linear models on the basis of DVI and DNSIb had the best estimating precision. This study demonstrated that it is feasible to estimate quantitatively the SL of cotton disease using the spectral indices of TM satellite image.