This paper proposes a spectral-based image daltonization algorithm for the dichromats getting the best visibility against their color deficiencies. The proposed model extracts the visible and invisible spectra to dichromatic vsion based on the projection theory of spectral space to/from 2-D dichromatic cone space.. First, the fundamental spectra C* visible to the trichromats are captured from the conventional sRGB camera images by a pseudo-inverse projection.. Second, the fundamental spectra C*_DIC visible to the dichromats are obtained by operating the extended matrix R_DIC based on matrix-R theory. Thirdly, the lost spectra ΔC_DIC are calculated as the difference between C* and C*_DIC.The key point in this paper is to make use of the lost spectra for image daltonization revived again. Though the lost spectra ΔC_DIC are invisible to dichromats if left alone, they are shifted into the visible spectral region and added to the the fundamental spectra C* of source image. As a result, the dichromatic image visibility is dramatically improved.The optimal spectral shift is determined to maximizing the spectral visibility for the dichromats and minimizing the visual gap from the normals. The proposed algorithm is designed to solve a contradictory demand to cope with both dichromats and normals.The paper shows how the model works well for image daltonization better than the existing popular color blind simulators.