Bangladeshi style: A way of facial artistic stylization in visual effects

摘要

In computer graphics field, stroke-based painterly rendering (SBPR) is the program to convert photographs into paintings, such as oil painting, watercolor and sketches. Beyond specific techniques to render real photographs, major challenge is how to understand real photographs as human visual perception and cognition according to artists' painting motivation, which parts should be drawn and how to draw strokes. Before starting to draw with pigments, real artists commonly draw a sketch using simple lines to mark out final stroke layouts which navigate to control stroke drawing trajectories and poses. This can be treated as a blue print in their mind. In SBPR, it is also a powerful tool to guide strokes configuration. In this paper, we propose a image-based method to extract this blue print called guidelines (GL). We generate GL by combining color gradient and morphological structure of real photographs. In the experiment, we apply our GL into Bangladeshi traditional mask automatic animation generation with webcam video. The results show that our method is sufficient and efficient to assist SBPR generation.