In color vision application, for the images with multiple colors, object can not be clearly separated from complicated background. The discrimination of the surfaces can be enhanced by selecting appropriate wavelength intervals of illumination. Firstly, the reflectance functions of all the surfaces were calibrated by four standard references. Then partial least squares method was used for selecting the effective wavelength interval of illumination. The variables important in projection (VIP) scores of wavelength intervals were considered as selection criteria. Wavelength intervals with the VIP > 1.0 were selected for illumination. Finally, three effective wavelength intervals of LED illumination were selected to separate all the surfaces of experiment image simultaneously. For separating all the surfaces of experiment image simultaneously and improving discrimination, the experiment was carried out. The experiment results demonstrate the usefulness of this method.