Karakterizacija kolorimetrijskih vrednosti otisaka štampanih goniohromatskim pigmentima

Abstract

U okviru disertacije predloženo je rešenje za određivanje kolorimetrijskih vrednosti otisaka štampanih goniohromatskim pigmentima na osnovu odziva digitalne kamere. Predmet rada predstavljao je razvoj modela karakterizacije kamere prilagođenog za fitovanje više-ugaonih podataka, kao i ispitivanje uticaja parametara predložene metodologije na tačnost procene vrednosti boja kamerom. Razvijeni model, baziran na veštačkim neuronskim mrežama, omogućio je postizanje zadovoljavajuće preciznosti merenja boja, procenu vrednosti boja svih testiranih mernih geometrija na osnovu snimaka u jednom mernom uglu, a pokazao je i visok stepen adaptivnosti na promenu osvetljenja koje se prilikom merenja koristi. Model je optimizovan primenom genetskog algoritma, čime je njegova efikasnost znatno unapređena.

The thesis proposes а solution for colorimetric characterization of prints enhanced with goniochromatic pigments by means of a digital camera. The subject of the research was the development of a camera characterization model adapted to fit multi-angular data and the assessment of the proposed framework parameters impact on the accuracy of camera-based color measurement. The developed model, based on artificial neural networks, enabled accurate color measurement with a satisfactory level of accuracy, estimation of color values of all analyzed measurement geometries on the basis of images obtained in one detection angle, and was proved to be very adaptive to the change of the illuminant used during the measurement. The model was optimized by means of a genetic algorithm, which led to the significant improvement of its efficiency.