We presents a method for automatic creation of a three-dimensional model from a traditional cel animation with a rotating camera around the character. Using the proposed system, children or amateurs may easily create a three-dimensional model with a little effort of finding a proper input animation.
Various computer vision techniques [e.g. Laurentini] are proposed for creating three-dimensional models from image sequence. However, these methods do not work for cel animations since characters in cel animations are not the exact projections of valid three-dimensional geometry. In contrast, our method creates a model that is valid in the topological sense by assuming that the character is an articulated model consisting of some components, which can be detected from its silhouettes.
Copyright (c) 2003 Yasunari Hashimoto