Not only spatial noise but also spectral noise exists in hyperspectral images.Only spatial noise but no spectral noise is removed with traditional image filters.To improve this situation DSGF(Derivative based Savitzky-Golay Filter) method was presented here.First noise level of each especial band is determined based on the second derivative of reflectance,then reflectance is filtered twice by Savitzky-Golay filter sizes,which results in removing spectral noise in hyperspectral image.In the meanwhile most fine features of reflectance are remained after hyperspectral image is filtered pixel by pixel by DSGF filter.