Quantified Evaluation of Various Spectral Encoding Techniques

Abstract

The goal of spectral encoding is to provide a limited dimensional space that preserves as much information about the original reflectance spectrum as possible. Recently, several spectral encoding techniques have been suggested by Nakaya and Ohta, as a simplified improvement over the LabPQR spectral encoding technique. To directly compare the reproduction accuracy between the new techniques and LabPQR a technique was developed to quantify the accuracy of three spectral encoding algorithms. Thousands of random spectra were generated and converted using the three spectral encoding techniques into six dimensions. The six spectral encoded values were then used to regenerate the target spectra. Simple RMS difference, Delta E 2000, and Delta E 2000 under a different illumination was used to evaluate reproduction quality. It was found that the LabPQR technique maintains higher reproduction capabilities than the alternate techniques.