Research On Spectral-Based Multiple-Ink Reproduction And Its Implementation

Conventional four-color printing systems often use the metamerism phenomenon to obtain the aim of chromatic reproduction. That is, color matches under a specific observer and illuminant are often changed under other observers or illuminants. The critical reason for the phenomenon lies in that conventional printing performs color matches according to the tri-stimulus value but spectral reflectance. With the continuous developments and advancements of modern printing technology, the people’s requirements of reproduction is higher and higher, high-fidelity printing has become a new industry standard. High fidelity means multiple-ink, which aims to spectrum matches. Thus it can make reproduction always keep consistent with manuscript when viewed under any environment.The research paper is concerned with the spectral-based multiple-ink reproduction. Its goal is to minimize the metamerism between the original manuscript and its reproduction. In order to achieve the purpose and thus eventually obtain the best spectrum matches, firstly the set of predicted primary colors for printing corresponding to the manuscript is needed, but not any more the traditional yellow, magenta, cyan and black colors are directly used. Secondly the corresponding spectrum prediction model is builded on the basis of the above primary colors for printing, which is the theoretical foundation of color separation. The whole flow must aim to the minimum spectrum errors including root mean square error of spectrum and metamerism index. With surrounding the frame, this paper has mainly carried out the following research works:1. Aiming at that the spectral-based multiple-ink reproduction does not use only traditional yellow, magenta, cyan and black colors as primitive colors for printing, color component prediction is needed to obtain the set of primary colors corresponding with the manuscript, which can be spectral reconstructed by the linear combinations of the above set of primary colors. The paper proposed the means of rotating principal component analysis with the finally gained eigenvectors representing the predicted set of primary colors, which adopts the maximum variance-based quadrate rotating with the eigenvectors gained from conventional principal component analysis. It aims to make the components in the corresponding directions have the maximum variance contributions to the original spectral reflectance while keeping eigenvector space quadrature, which can promise the irrelevance of primary-color and its spectrum reconstruction accuracy.2. The means of principal component analysis can not promise the independency of predicted components, which means that each primary color can not be gained from the combination of other primary colors in practical printing, and can only predict little primary color component, so this paper also adopts the means of independent component analysis, which aims to get a group of better predicted components representing real inks. But conventional independent component analysis is limited to the high computing complication and the slow convergence rate, and the improved independent component analysis without computing the derivative of matrix and its inversion iteratively is limited to the unstability because of the relationship with the selection of original values. Thus this paper proposed the damped modified independent component analysis introducing a dampness coefficient in the improved independent component analysis, which can promise the resulting convergence.3. Because the color components predicted through the above means of statistical analysis often can not completely match with the actually used primary colors, only can the optimal ink combination be selected as the actual primary-color for the spectral-based reproduction, which can promise the matching with the original predicted primary-color to the maximum extent. This paper proposed an optimal ink selection based on the measuring with mixed distances, which gives attention to the change of color lightness and the character of spectrum shape.4. After selecting the actually used set of primary colors for printing, the relationship between theoretical dot area percentage and spectral reflectance can be achieved through the building of spectrum prediction model. This model is based on the six-ink in theory, but eventually four inks (three colors plus black ink) are used because of economic cost, ink amount limitation and computing complication while real reproducing. Thus the paper proposed the spectrum prediction model based on the four-ink dot gains, which not only takes into account the direct impact of each ink superposition on the dot gain of every single ink, but endows each impacting factor with the corresponding weight coefficient based on the theory of Demichel probability equation. Eventually the effective dot area prediction of every single ink under multiple-ink superposition is computed by the weighted sum of impacting factors and weight coefficients, and then the spectrum reflectance of the corresponding color in superposition is computed by the Demichel probability equation and four-ink YNSN model.5. The application of the spectrum prediction model includes two facets with forward and backward directions. The forward directed operation is to predict the spectral reflectance of corresponding superposed color by the theoretical dot area percentages, which is the theoretical basis of the coming backward directed operation. The application of backward direction is to predict the theoretical dot area of every included single ink on the basis of the spectrum reflectance of some specific sample, and evaluate whether the above theoretical dot areas meet the precision requirements of spectral-based reproduction through the application of forward direction. The application of backward direction is essentially color separation. The color separation in this paper includes the selection of optimal four-ink combination and the theoretical dot area prediction of every single ink under the corresponding ink combination. Furthermore, the paper proposed a genetic algorithm based on the self-adapting choosing strategy in order to achieve the theoretical dot area prediction with global optimization. Lastly the author carried out the real output test on the proofing device-Canon iPF5100according to the above color separation results, and compared the output spectrum and the original spectrum one by one in order to verify the performance of the spectrum prediction model.