A colour harmony model for two‐colour combinations

Color harmony for image indexing

“…mony scores for images containing a lot of different colors are often low, which is rather obvious since the probability to find a disharmonious color combination usually increases with the number of colors. It is not surprising that the results agree with the two-color harmony principles given by Ou & Luo [16].…”

“…Research in harmony has usually been carried out on rather restricted combinations of colors, like two-color combinations. An extensive study by Ou & Luo [16] investigates harmony in two-color combinations in order to develop a quantitative model for prediction. In psychophysical experiments, color pairs were assessed on a category scale related to harmony.…”

“…For deriving color harmony scores, we utilize the quantitative two-color harmony model developed by Ou & Luo [16]. In psychophysical experiments, observers were presented with color pairs, and harmony scores were assessed for each color pair using category scaling.…”

“…To compensate for the "rubber band effect" (observers are using the scale in different ways), we calibrate observers to a common scale by adjusting judgments until they have the same mean and variance. To measure the inter-observer agreement, a measure of how well observers agree with each other, we adopt the method and terminology used by Ou & Luo [16]. The agreement can be derived by averaging the Root Mean Square (RMS), between each observer's judgment and the overall mean, given by …”

Erratum

“…Colour harmony is schemed to satisfy human response for two 28 or more juxtaposed colours 29 (e.g., apply complementary, analogous, triadic rules and so on in the colour wheel). Therefore, it is more reasonable to consider the colour harmony as a relative comparison in a reference harmony distribution, rather than an absolute measurement.…”