Nowadays pictorial images are more often shown on displays than on paper. Therefore, displaymanufacturers have been trying to improve the image quality of their displays to increase theirmarket share. To improve the image quality, not only good hardware technology and imageprocessing algorithms but also good color difference equations are needed to predict the overallcolor difference between pictorial images shown on different panels or manipulated using differentimage processing algorithms etc. Color difference equations are also developed for the use ofindustrial purposes. However they were developed in limits of surround condition and applications. The purposes of this research is to clarify the effect of surround condition and magnitude of colordifference on perceptual color difference for complex image. The experiment under four surroundconditions was carried out to achieve these purposes. The collected data by psychophysicalexperiment was used to develop image color difference metric under various surround condition.Before conducting the main experiment, pilot test was conducted to investigate the effect of colordifference magnitude on perceptual image difference. Pilot experiment tested the performance of each conventional color difference equation such asCIE ΔE*ab, CMC(l:c) and CIEDE2000 while the psychophysical experiment including wide rangeof color difference stimuli was conducted to evaluate perceptual color difference between originalimage and manipulated image. Twenty observers were participated in the experiment and 195stimuli were used for the magnitude estimation. Main Experiment was investigated how perceptual color difference is shifted by changing surroundluminance level and magnitude of color difference. There are four surround conditions, dark, dim,average and bright. 996 stimuli were prepared for the experiment. 500 randomly selected stimuliwere used for each surround condition. Twenty-three observers were participated in the mainexperiment. They were asked to evaluate the perceived color difference between original andmanipulated images with magnitude estimation method.
Publisher
Ulsan National Institute of Science and Technology (UNIST)