Image Colorization with an Affective Word

Abstract

An important role of image color is the conveyer of emotions (through color themes). The colorization is less useful with an undesired color theme, even semantically correct, which has been rarely considered previously. In this paper, we propose a complete system for the image colorization with an affective word. We only need users to assist object segmentation along with text labels and give an affective word. First, the text labels along with other object characters are jointly used to filter the internet images to give each object a set of semantically correct reference images. Second, we select a set of color themes according to the affective word based on art theories. With these themes, a generic algorithm is adopted to select the best reference for each object. Finally, we propose a hybrid texture synthesis approach to colorize each object. Our experiments show that the results of our system have both the correct semantics and the desired emotions.