カラー画像からのアニメーションキャラクタの同定

抄録

The goal of this paper is to provide a method to identify animation characters from color images, in order to automate image tagging for web artist communities.The solution will take the form of a supervised or semi-supervised learning algorithm based on tools from graph theory and kernel methods, and extending previous results for real human and animation images. Segmentation and classification techniques leveraging specifics of animation images will be designed for this purpose.Such a method will be expected to perform well on a data set of animation character images without background.