Barvanje črno-belih fotografij z modrostjo množic

Weiss, Amela (author), Demšar, Janez (mentor)

Starejše fotografije obstajajo le v črno-beli izvedbi. Neobstoječe barve takih fotografij tako predstavljajo delčke izgubljene zgodovine. V magistrskem delu se ukvarjamo z barvanjem črno-belih fotografij. Fotografije smo barvali z modrostjo množic, kar je nov pristop. Izdelali smo merski instrumentarij (spletno aplikacijo) in izvedli raziskavo na udeležencih. Pri računanju razdalj med barvami smo uporabljali matematične modele, ki najbolj ustrezajo človeškemu dojemanju barv. Rezultate množice smo primerjali z rezultati algoritma, poiskali kategorične razlike pri udeležencih in pogledali, kako se zanesljivost rezultata množic spreminja s številom udeležencev. Pokazali smo, da z večjim številom udeležencev napaka vedno bolj konvergira k neki vrednosti. Moški in ženske barvajo približno enako dobro, vendar so izbire moških bolj razpršene. Rezultate množice smo primerjali z rezultati algoritma za barvanje črno-belih fotografij, ki bazira na globokih nevronskih mrežah. V primerjavi z algoritmom so ljudje boljše reševali naloge s kontekstom. Nova metoda barvanja črno-belih fotografij z modrostjo množic izgleda obetavna, vendar so potrebne dodatne raziskave na večjem številu ljudi.

• Language: Slovenian
• Keywords: barvanje
• Work type: Master's thesis/paper
• Typology: 2.09 - Master's Thesis
• Organization: PEF - Faculty of Education
• Year: 2020
• PID: 20.500.12556/RUL-117178
• COBISS.SI-ID: 20993027
• Publication date in RUL: 08.07.2020

WEISS, Amela, 2020, Barvanje črno-belih fotografij z modrostjo množic [online]. Master’s thesis. [Accessed 14 August 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=117178

Secondary language

• Language: English
• Title: Colorizing grayscale photographs with the wisdom of crowds

Abstract

Older photos only exist in black and white. The non-existent colors of such photographs thus represent fragments of lost history. The master's thesis deals with the coloring of black-and-white photographs. We colorized the photos with the wisdom of the crowds, which is a new approach. We created a measuring instrument (web application) and conducted a survey on the participants. In calculating the distances between colors, we used mathematical models that are most relevant to human perception of colors. We compared the results of the crowd with the results of the algorithm, looked for categorical differences in participants, and looked at how the reliability of the result of the crowds varies with the number of participants. We showed that with more participants, the error converges to some value. Men and women pick color about equally well, but men's choices are more diffuse. We compared the results of the crowd with the results of an algorithm based on deep neural networks. Compared to the algorithm, people were solving better tasks where contextual knowledge is needed. The new method of coloring black-and-white photographs with the wisdom of the crowds looks promising, but more research is needed.

• Keywords: coloring