This thesis presents a non-photorealistic rendering technique to create focus-directed stylization using RGBD data, which contains not only RGB color information but also depth values. This thesis applies depth data into video stylization. Our method consists of four parts, line drawing, image abstraction, focus-directed framework and focus point tracking. In the first part, our approach tries to extract more lines from depth as well as RGB color information. For the second part, we develop an adaptive trilateral filter method to produce detail preserving stylized images that show clear depth cue. For the third part, a focus-directed framework is designed to produce a DOF-like effect based on a focus point which is selected by a user. Furthermore, the focus point is tracked through the video to get temporal coherence.
