Research has been initiated to determine a set of six basis colorants which are the best representation of artwork such as paintings. That is, their spectral information can be accurately reconstructed by linear combinations of the six estimated colorants. Since each painting is possibly created by different colorants, the six estimated colorants are image dependent. The clue leading to the six estimated colorants is the six eigenvectors determined from the corresponding spectral measurements. The relationship between the six eigenvectors and estimated colorants is merely the linear transformation (or geometrical rotation). Based on this faith, a constrained-rotation engine using MATLAB was devised to perform the transformation from the eigenvectors to a set of all-positive vectors as the estimated colorants. Once a set of reasonable colorants is uncovered, this set of colorants can be used to synthesize the original artwork with the least metameric effect between the reproductions and originals.