Research on Spectral Reconstruction Accuracy of Color Reproduction Based on PAC

Abstract

The key of spectrum color reproduction is to study the spectral information of original images, and reconstruct the spectral curve of the target color. The accuracy of reconstructed spectra is influenced by many factors. In order to study the influence factors of reconstruction accuracy, two kinds of color cards Munsell Color Matt and Color Checker Classic were selected as the spectral reflectance data samples, two kinds of linear dimension reduction models of PCA were established and different number basis vectors were selected to separately reconstruct the spectrum. Then the influence of the different models and the basis vector numbers on the reconstruction accuracy was evaluated. The experimental results showed that the accuracy by model one was better than that of model two in RMSE GFC and color difference. In two kinds of color cards, when the number of basis vectors obtained by model one reached 6, the color difference was less than 1, the RMSE was less than 0.01; the GFC was up to 0.999. So the optimal scheme of reconstructing spectral images is to select spectral dimension reduction model one and 6 basis vectors.