A Cartoon Image Annotation and Retrieval System Supporting Fast Cartoon Making

摘要

Cartoon image annotation and retrieval is critical for cartoonists to effectively and efficiently make cartoons by reusing existing cartoon data. To successfully achieve these tasks, it is essential to build an abundant cartoon database first and then annotate and retrieve them efficiently. This paper studies and realizes a cartoon images annotation and retrieval system which supports interactive fast cartoon making. Firstly, the framework of the system, which consists of three modules including cartoon database management, cartoon image annotation and cartoon image retrieval, is introduced. Secondly, considering the features of cartoon scene images having less color number, higher hue saturation and wider color distribution than nature scene images, a cartoon scene annotation method based on the idea of retrieval and adopts image matching and label clustering method is proposed. To further enhance the image matching performance, global color feature and local CSIFT feature are combined in this method. Thirdly, as for the cartoon character image, which is another kind of important reusable image, a shape-based retrieval method is put forward. Here, shape feature extraction is focused to make sure that all the chromatic and monochrome character images can be annotated. Finally, various browsing and retrieval methods based on the feature of color, local invariant feature and shape, etc., which contributes a lot to the reuse of cartoon images, are proposed. The established system has already been applied to the cartoon enterprise and has been proved to be effective in supporting fast cartoon making.