Color-based semantic characterization of cartoons

Abstract

In this paper a statistical color-based approach for semantic characterization of animation movies is proposed. As color is a major feature of animation movies (each movie has its own color distribution) it should be the ideal tool to use in deriving semantic information about the used color artistry concepts or about the sensation Induced by the movie's color distribution. First, the movie is splitted into shots by detecting the video transitions. A movie abstract is then automatically generated by using certain key frames for each shot color reduction is applied on each key frame using an error diffusion approach on an predefined color palette (having each color already named). Then, a global weighted color histogram of the movie is computed by taking into account each shot relevance, to serve as a basis of computation for relevant color-based statistics, namely: pure color distribution, light/dark color ratio, hard/weak color ratio, warm/cold color ratio, complementary colors ratio, adjacent colors ratio and color diversity ratio. These parameters are then used to derive meaningful color-based semantic information. The proposed approach was tested on several animation movies.