Data-Driven NPR Illustrations of Natural Flows in Oriental Painting

The strokes of rivers and falls are important elements in oriental paintings to express surface sprays, smooth water lines, and coherent diffusion. However, it is not trivial for general users to draw water flows with artistic styles even with a commercial software to generate oriental painting animations. Aiming at this point, this study designs a data-driven system to extract arbitrary painting styles from an existing oriental painting, animate these stylizing strokes, and transfer the styles to other paintings. Our system extracts an initial flow pattern by analyzing the structure, placement density, and ink density of strokes and automatically computes a flow field according to water boundaries and flow obstacles. The flow field is generated by solving the Naiver-Stokes equations in real time and its painting style is also extracted as patterns of strokes with their location, oscillation style, brush pattern, and ink density. Finally, the extracted strokes are dynamically generated and animated with the constructed field with controllable smoothness and temporal coherence. Furthermore, our system can transfer the extracted painting styles to animate the water flow of another oriental paintings. The overall flow animation is pleasant and delivers the spirit of the existing painting without flickering artifacts commonly existing in a stroke-based non-hotorealistic rendering (NPR) animation.