The investigation of effects of digital proofing systems used in colour management on print quality with neural networks

Assessing print quality by machine in offset colour printing

“…Various factors affect print quality: paper (ink demand and ink pigment penetration increase with increasing surface roughness, for example [2]), ink, printing press, digital proofing system used in colour management [3], ink-water balance [4], operator actions, and the quality of printing plates. A large variety of deficiencies and problems may appear during printing: misregistration of printing plates, paper wrinkle formation [5], poorly formed paper surface [6], too high linting due to poorly formed paper surface [7], piling, mottling, web breaks, etc.…”

Colour management of a low‐cost four‐colour ink‐jet printing system on textiles

“…The LUT is sometimes prepared by a printing house based on the results of their own products. One of the disadvantages of this method is the work required to print and measure the large number of samples [8–10]. It has been suggested that, for the characterisation of printers, the number of samples required is at least 200 [11].…”

“…The multilayer perceptron (MLP) is a type of feedforward ANN that is utilised extensively in the colour management system in conjunction with the back‐propagation learning algorithm [9,15,22]. However, some faster ANN algorithms, such as conjugate gradient, quasi‐Newton and Levenberg–Marquardt (LM), which use standard numerical optimisation techniques, have been evaluated for print quality control [8]. The algorithm of LM is a kind of approximate Newton’s method, which uses the second‐order derivatives of the cost function to obtain better convergence behaviour [8,23].…”

“…However, some faster ANN algorithms, such as conjugate gradient, quasi‐Newton and Levenberg–Marquardt (LM), which use standard numerical optimisation techniques, have been evaluated for print quality control [8]. The algorithm of LM is a kind of approximate Newton’s method, which uses the second‐order derivatives of the cost function to obtain better convergence behaviour [8,23]. The number of samples required for an effective neural–network characterisation model is relatively high and the extra work required may not justify their use over conventional polynomial methods [24].…”

Use of Colour Management to Achieve Matching of Prints on Cotton Fabric with Simulation on Paper

“…Barvni pro l je opis barvnih lastnosti naprave, za njegovo izdelavo pa se uporabljajo neugebuerjeve enačbe, polinomske funkcije, nelinearne pretvorbe, interpolacija z uporabo LUT tabel [13] in umetne nevronske mreže ANN (angl. arti cial neural network) [14]. Danes so tovrstni načini del programske opreme za izdelavo barvnih pro lov.…”