Analysis of observer and cultural variability while generating preferred color reproductions of pictorial images

The goal of color reproduction is generally to maximize the accuracy between an original scene or image and a reproduction. The focus of this paper is to better understand color preference reproduction where the original image is the cognitive representation of the reproduced image rather than a physical original. In order to quantify the degree of accuracy when measuring cognition, the implementation of psychophysical techniques is required. This paper outlines two psychophysical experiments, a ranked-order design utilizing hard-copy reproduction, and the method of adjustments utilizing soft-copy reproduction. Both experiments depend on the application of five independent colorimetric adjustment dimensions to manipulate a set of scenic images so that observers could generate their preferred version of the given image. Sources of preference variability such as observers' cultural background, image content, and inter- and intra-observer variability were evaluated from the observations made in the experiments. The experiments yielded that inter-observer variability is approximately twice the magnitude of intra-observer variability. When analyzing the data set for image content variability, images that contained people (faces) were routinely less variable than images without people (faces). Furthermore, despite the fact that preference variability due to observers' cultural background was found to be statistically significant, it was also demonstrated not to be visually significant. Finally, variability in preference due to image content and differences among observers were visually more significant than the variability due to cultural background. In order to demonstrate the amount of variability experienced in these experiments, comparison image sets were generated and simulated sets are included in the analysis in this article.