The spectral features and differences in bright vegetation area, shaded vegetation area and water area were investigated by the experimental data from four medium resolution remote sensing images of ALOS AVNIR-2, CBERS-02B CCD, HJ1A-CCD2 and Landsat 7 ETM. Based on the near-infrared band and normalized difference vegetation Index (NDVI), Normalized Shaded Vegetation Index (NSVI) was constructed and the enhancements of spectral differences and classification effect were also evaluated. The results show that NSVI has increased the relative diferences of the spectra in bright vegetation area, shaded vegetation area and water area, and reduced probability of misapplication for the spectral data. The NSVI threshold method was employed to classify the four experimental images. The overall accuracy is over 97%, and the overall Kappa coefficient is above 0.96. The detection accuracy of the shaded vegetation area is over 94% and the Kappa coefficient is also higher than 0.96. By using radiation differences of the near-infrared band between the ground objects, NSVI can solve the problem that NDVI can only partially weaken the topographic effect and enlarge the spectral differences among the ground objects. NSVI enhances the validity of the ground objects especially in the shadow detection and avoids the “saturation” problem of NDVI. It can provide a new solution to remove the shadow in remote sensing images.