Characterization of High-Fidelity Color Printing Devices Using Illuminant-Independent Approaches for Color Imaging Application

Abstract

The aim of this research was, optimally to derive a multi-spectrally illuminant-independent type of High-Fidelity (Hi-Fi) multi-color printing device characterization model, which could reconstruct spectra of every color concerned in high-dynamic-range of original images for an accurate color-matching or - reproduction application. In this study, the Adobe RGB color space under the D50 illuminant condition was used as the reference working space. To have the ability to adapt the change of light sources, two works were preliminarily carried out. Firstly, an ideal type of CIEXYZ camera characterization model was derived using a spectral reconstruct method, based on basic vector under the D50 condition. Via the estimation and the reconstruction of the spectrum of every pixel in question, the XYZ under any light source considered could be obtained accordingly. Therefore, any Adobe RGB format of Hi-Fi complex color images could be transformed into the multispectral type of images via the CIEXYZ camera model derived. Secondarily, the characterization of 7-ink CMYKRGB printing process, using FM screening technique, was implemented and tested in terms of single stimulus. The superset of CMYKRGB i.e. Hi-fi color set) was divided into seven 4-ink groupings, including KGCB, KBMR, KRYG, RKYM, GKYC, BKMC, and also CMYK. Each subgroup contains three chromatic inks and one black ink. Both the multispectral (MS, i.e. narrowband) and the broadband (BB) types of characterization models, using 3rd-order and 2nd-order polynomial regression equations respectively, were developed in this stage. They all applied singular value decomposition method (SVD) and constructed including both transforms of forward (i.e. fractional dot area (FDA) to XYZ/L*a*b*) and reverse (i.e. XYZ/L*a*b* to FDA) processes for each subgroup. The transform between device-independent data and device-independent data was derived and defined via measuring a number of colors in the IT8.7/4 training/test target produced using each 4-ink grouping subset. An approach of key component replacement (KCR), which is similar to GCR (gray component replacement) applied in traditional CMYK printing process, was integrated to implement the BB type of a multi-ink color separation algorithm. Additionally, in order to reconstruct the spectra and adaptively map to the corresponding tone values (fractional dot area) for every color in test/training datasets, the MS approach in the reverse process was optimized by iterating the KCR components of BB algorithm in terms of the measure of RMSE (root of mean square error). The predictive performances of 2 models derived were tested for each of subgroup in the IT8.7/4 datasets mentioned above. Two measures of both the mean CIEDE2000 (i.e. E*00) and the mean RMSE were used. The results showed that two types of algorithms, used in the printing process of 7-ink CMYKRGB, were successfully proposed. The mean E*00 values are 1.04 and 0.79 in the forward and the reverse processes respectively for the MS type. Also 0.00098 of the RMSE value was obtained for the reverse transform of the MS type. It implies the spectral reflectance of every color in datasets was satisfactorily reconstructed. Subsequently, following the preliminary study, by taking two matters of “metamerism” and “wide-gamuts” in mind, feasible application modules of Hi-Fi multi-color process toward the spectral/complex color images was derived and proposed. Then, Sets of image-processing algorithms were implemented in the process of scanner-to-monitor, and then monitor-to-printer color transform in the study. Eventually, to cross verify performances of models derived, a set of forced-choice paired comparison psychophysical experiments carried out under a viewing phase of CIE D50 illuminant.