Printer color management using pareto-optimization and efficient calibration techniques

Color management for the printing of digital images is a challenging task, due primarily to nonlinear ink-mixing behavior and the presence of redundant solutions for print devices with more than three inks. In this thesis, Pareto-optimal formulations governing L*a*b*-to-CMYK conversion for color printing are presented. The Pareto-optimal method offers user control over trade-off problems such as cost versus reproduction accuracy, allowing for user-specified print objectives and the use of constraints such as maximum allowable ink and maximum allowable error. In the case of out-of-gamut colors, clipping to the closest in-gamut color is achieved by minimizing colorimetric error, $DE*ab$.; Two implementations of the Pareto-optimal approach are presented: the program NeuralColor, and the program OptInterpol. These schemes operate using a 150-color characterization set selected to capture efficiently the entire CMYK gamut. The first implementation uses artificial neural networks as transfer functions between the L*a*b* and CMYK color spaces. The second scheme is based on a reformulation of tetrahedral interpolation as an optimization problem. Both approaches yielded prints with errors less than three units in the L*a*b* space for the case of minimization of error.; The problem of controlling grey-component replacement (GCR) in a printed image is addressed through a multiple-step approach based on the Pareto-optimal method. The GCR formulation developed in this thesis generalizes solution methods for specifying arbitrary GCR as bounded by the minimum- and maximum-GCR solutions, finding solutions that use only chromatic inks, and obtaining solutions that use at most two chromatic inks and black ink. Prints obtained using NeuralColor are accurate within two to four units in the L*a*b* space across all levels of GCR.; Finally, methods for maintaining accurate printer characterization are presented. The model errors that result from a change of paper or a change of printer cartridge are modeled using linear polynomials, quadratic polynomials, and artificial neural networks. Additionally, a method is investigated that updates a previously existing artificial-neural-network printer model by adding heavily weighted recalibration data, to the original characterization set. A significant reduction in error was realized by incorporating these techniques into the color-management program NeuralColor. The most successful of these methods was a quadratic polynomial correction model, which removed 90%, of the error-introduced by a change of paper stock, and all of the error introduced by a change in toner cartridge. A general conclusion regarding recalibration methods is that simple models exhibiting global control are preferred over more complex models which exhibit local control.