AKAIKE INFORMATION CRITERION IN THE EDGE ANALYSIS OF THE SCREEN ELEMENT

The analysis of the edge of the screen element is based on the selection of the linear sample on the paper-ink system in comparing models and reality. Although Yule and Nielsen suggested the Gaussian distribution for the description of the line spread function, our model, based on the stochastic approach of subsurface light scattering in paper, gives the complete description with the Lorentzian distribution. To determine which of the two proposed models, Gaussian or Lorentzian, gives a better approximation with respect to the measured data, Akaike information criterion has been used. As the observed profiles are asymmetric, both edges have been analyzed. It was not possible to distinguish which model better describes the resulting measurements, because of the extremely high value of the correlation coefficient for both models. Therefore the AIC method was applied using the routines in Origin 8.5 that was used to analyze the measured data. The Akaike weight demonstrates that the Lorentzian model better describes the LSF than the Gaussian model.Original Abstract: Analiza ruba rasterskog elementa temelji se na odabiru otisnutog linijskog uzorka na sustavu papir-boja i usporedbi modela i stvarnosti. Iako su Yule i Nielsen predlozili Gaussovu raspodjelu za opis funkcije razmazivanja linije, nas model, temeljen na stohastickom pristupu podpovrsinskog rasprsenja svjetlosti u papiru, daje cjelovit opis pomocu Lorentzove raspodjele. Da bi utvrdili koji od dva predlozena modela, Gaussov ili Lorentzov, daje bolju aproksimaciju s obzirom na mjerene podatke, bio je koristen Akaikeov kriterij informacije. Kako su promatrani profili asimetricni, analizirana su oba ruba. Pri tome nije bilo moguce razluicti koji model bolje opisuje rezultate eksperimenta, zbog iznimno visoke vrijednosti koeficijenta korelacije za oba modela. Stoga je koristena AIC metoda pomocu rutine u programu Origin 8.5 koji je koristen za analizu izmjerenih podataka. Akaikeova tezina pokazuje da Lorentzov model bolje opisuje LSF od Gaussovog modela.