The infrared thermal images and leaf water potential (ψL) of two cotton cultivars named Xinluzao-33 and -13 were obtained at five growth stages after five different irrigation processes. The temperature values of the canopy sunlit leaves were extracted from the thermal images and the crop water stress indexes (CWSI) were calculated. The analysis result showed that the canopy CWSI and ψL of two cotton cultivars at five different irrigation levels changed inversely with the growth stage, and the ψL was consistent with the growth of cotton. Their change orders were all W5> W4> W3> W2> W1. The ψL of the tested samples was estimated by using the linear regression function between the CWSI and ψL of the modeled samples ($r=0.8903**, n=24, α=1%). The measured ψL had a significant linear correlation with the estimated ψL ($r=08971**,RMSE=0.1223, n=24,α=1%). The estimated ψL had a relative error of 0.0981. Its estimation accuracy was 90.2%. The result showed that the CWSI based on the infrared thermography could be used to estimate the leaf water potential of cotton rapidly and nondestructively in real time and could be used to predict the water potential accurately.