POPRAVLJANJE BARV NA SLIKAH S POMOČJO BARVNEGA KALIBRATORJA IN ALGORITMOV BARVNE VZTRAJNOSTI

• Main Author: Črešnar, Štefan
• Other Authors: Zazula, Damjan
• Format: Bachelor Thesis
• Language: Slovenian
• Published: Š. Črešnar 2016
• Subjects: barvni kalibrator; digitalna obdelava slik; popravljanje barv; algoritem barvne vztrajnosti; algoritem za nastavitev beline; colour calibrator; digital image processing; colour correction; colour constancy algorithms; grey world assumption; info:eu-repo/classification/udc/621.397.122(043.2); sami
• Online Access: https://dk.um.si/IzpisGradiva.php?id=60411; https://dk.um.si/Dokument.php?id=100394&dn=; http://www.cobiss.si/scripts/cobiss?command=DISPLAY&base=cobib&rid=19918870&fmt=11

V diplomskem delu opisujemo in preverjamo postopke barvne vztrajnosti za popravljanje barv na digitalnih slikah s pomočjo barvnega kalibratorja in algoritma povprečne sivine. Na slikah naš algoritem samodejno prepozna barvni kalibrator. Tega uporabimo za prilagajanje barv z minimalno kvadratno napako in za izračun uspešnosti. Pri algoritmu povprečne sivine s kalibratorjem samo preverjamo uspešnost. Izvedli smo poskuse z dvema naboroma slik. Enega smo izbrali iz spletnih baz, drugi sklop smo posneli sami v laboratorijskem okolju. Slikali smo s standardnim digitalnim fotoaparatom na dva načina: enkrat skozi polprosojno ogledalo, ki ima prepustnost od 60 do 65 %, in enkrat neposredno. Sceno smo osvetljevali z različnimi svetlobnimi viri: dnevno svetlobo, studijskim reflektorjem, reflektorjem z žarilno nitko in zeleno svetlečo diodo. Analizo barvne vztrajnosti smo opravili v linearnem barvnem prostoru RGB. Poskusi so pokazali, da imata oba testirana algoritma podobno uspešnost. Prilagajanje s kalibratorjem se sicer v splošnem obnese nekoliko bolje, algoritem povprečne sivine pa je prepričljivo boljši na slikah, ki smo jih posneli v laboratorijskem okolju. A le na slikah, ki so posnete brez ogledala. Zadnji poskus na slikah, ki so posnete skozi polprosojno ogledalo, pa pokaže, da je na tem naboru slik algoritem za prilagajanje barv s kalibratorjem učinkovitejši. In diploma thesis, we describe and test colour constancy procedures for adjusting colours in digital images based on the coloured checkerboard and grey world assumption algorithms. Coloured checkerboard is automatically detected by our algorithm in images. After detection it is used for adjusting colours on the mean square error basis and assessing the success rate. For the grey world assumption algorithm, the checkerboard is only used to assess the success rate. Experiments were conducted on two sets of images. The first was selected from web databases and the second was taken in laboratory environment. We used standard digital camera and the images were captured in two different ways: once through a semi-transparent mirror with the transparency between 60 and 65% and once directly with no mirror. The scene was illuminated by different light sources: daylight, a softbox, a reflector bulb, and green lighting LED. Analyses were performed in the linear RGB colour space. Tests indicated that results of both algorithms were similar. In general, adjustment based on coloured checkerboard provided better results, although the grey world assumption algorithm really did better on images that were taken in the laboratory. But only on those that were taken without semi-transparent mirror. In the final test on images that were taken through semi-transparent mirror, the coloured-checkboard-based correction was more efficient.