Based on spectral indices method, this paper utilized the water content (Cw) and reflectance data of 67 fresh different type leaves from LOPEX' 93 database to establish the statistical model between leaf Cw and spectral indices at leaf level through 47 samples, and to test the model with the other 20 samples. The results suggested that fuel moisture content (FMC) and equivalent water thickness (EWT) as Cw demonstrators were different in reflectance spectral curves. The difference between FMC and EWT was large when they were utilized to retrieve the leaf Cw. The correlation coefficient between EWT and each spectral index was higher than FMC, but the forecast precision of FMC was better than that of EWT. The 7 spectral indices could all retrieve the leaf FMC accurately, but only the Ratio975, II and SR were suitable to estimate the leaf EWT. Spectral indices linear model on the strength of optimal subset regressions had the highest precision to retrieve the leaf Cw. Ratio975 might be the universal spectral index to estimate the leaf Cw. At canopy level, the simulated canopy spectra under different leaf area index (LAI) and Cw were derived from the PROSPECT and SAILH coupling models. In order to eliminate background influence and to precisely retrieve the Cw, soil-adjusted water index (SAWI) was proposed at the first time to indicate the information of near-infrared and short-wave infrared canopy reflectance. The ratio of SAWI and other spectral indices could dramatically eliminate the soil background, and effectively retrieve the vegetation Cw at canopy level. Spectral index (Ratio975 - 0.96)/(SAWI + 0.2) as improved Ratio975 could be used to compute the canopy Cw more precisely when LAI was ranging from 0.3 to 8.0 and Cw from 0.0001 to 0. 07cm.