Paintings can evoke emotions in viewers. In this paper, we propose a method for extracting emotions from paintings by using the colors that comprise the paintings. The proposed approach is based on a color image scale, which is one of the popular experimental scales focusing on the relation between colors and emotions. We first construct a color combination and emotional word dataset. To this end, we create a color spectrum from the input painting. We then search for the best matching color combination from the dataset, which is most similar to the color spectrum. The best matching color combination is mapped to the corresponding emotional word. Afterward, we extract the emotional word as the emotion evoked by the painting. To evaluate the proposed method, we compared the results of the proposed algorithm to those of a user study on the extraction of emotions from several paintings. Through several experiments, we show that the proposed method exhibits excellent performance with respect to predicting the emotions evoked by a painting. Finally, we propose an image exploration system based on the emotion extraction method mentioned above. In this system, users can explore painting images emotionally coherently.

A Color Image Scale is one of useful resources for allowing designer the expression of mood through the color combination. This study proposes a method for estimating the mood in Color Image Scale from three color combinations by using machine learning technique. After we find the correlation between the mood and the properties of color combination, we extract three dominant colors from an image. Finally, we estimate the mood of the painting by using the properties of three dominant colors.

No abstract