This paper proposes a spectral-based CVD (Color Vision Deficiency) model compatible with both of dichromacy and anomalous trichromacy. The spectral projection model based on Matrix-R extracts the lost spectra as a difference in the fundamentals between the normal and the color deficient. The lost spectra are re-used for image daltonization by an optimal spectral shift to maximize the spectral visibility or minimize the visual gap from the normal. The model rationally improves the scene visibility after daltonization. The proposed algorithm is designed based on the original key ideas of 1) Acquisition of fundamental C*_LMS (spectra visible to the normal) from sRGB camera image by a pseudo-inverse projection without expensive spectral image. 2) Foundation of projection matrix-R_CVD onto dichromatic and anomalous trichromatic spectral spaces by combining the cone responses in the table by DeMarco&Pokorny& Smith. 3) Extraction of fundamental C*_CVD (spectra visible to the dichromat or anomalous trichromat) by operating the matrix-R_CVD on the fundamental C*_LMS. 4) Introduction of complete OCS (Opponent-Color Space) to keep the perfect achromatic grayness in the opponent-color stage. 5) Estimation of lost spectra ΔC*_CVD as the difference between visible spectra C*_LMS to the normal and C*_CVD to the dichromacy or anomalous trichromacy. 6) Color blindness correction (daltonization) by reviving the lost spectra ΔC*_CVD with the optimal spectral shift into the visible waveband.