Toward a generic approach to stylization

摘要

This paper presents a first attempt toward a generic process for stylization. Guided by the scene content and constrained by the artistic or functional expectations of the end-user, our model produces, in real time, an expressive rendering of the scene based on the user settings. Our work is based on semantic description of shapes that allows logical inference on the scene content, performed via the description logic formalism. This inference yields semantic data that can guide the graphical restitution of the scene. This approach is illustrated by a practical application example that successfully stylizes a 3D scene in real time and for a predetermined style. The resulting stylization, even if style is selected in advance, automatically and successfully highlights different parts of the scene according to semantic data extracted for each object. The stylization results are then discussed before we conclude and present our future works.