Optimized spectral estimation methods for improved colorimetry with laser scanning systems

The colorimetric performance of a 3D laser scanning camera is predicted for a set of three sampling wavelengths (442 nm, 532 nm and 633 nm) currently in use. Methods of spectral estimation based on principal component analysis and on spline interpolation are considered. The average ΔE94 color difference over large sets of reflectance curves is used as a metric. Optimal sampling wavelengths are next investigated using unconstrained optimization of the average ΔE94 color difference for the set of reflectance curves from the full OSA-UCS catalog. Three up to six optimal sampling wavelengths are derived with both methods. The PCA method is found a hint better than the spline method overall, and the ΔE94 errors are reduced roughly by one half when the number of sampling wavelengths increases by one. Four or five sampling wavelengths become sufficient to bring the average error below the one just noticeable difference benchmark. Spectral estimation with 5 optimal wavelengths using the principal components derived from the OSA-UCS catalog appears adequate for the most demanding applications.