Abstract
In cartoon rendering of 3DCG, methods for generating a contour lines are classified into the method using the image processing or the method using the three-dimensional shape of the models. By selecting G-Buffers, the method using the image processing can accurately detect intended lines of users, but the lines are mechanical impression by the uniform thickness. The methods using the three-dimensional shape of the models can adjust thickness of the lines, but detection of the lines is dependent on the rendering algorithm and it can't detect the intended lines of users. In this paper, we separated contour lines process into detection and rendering. Our approach uses image processing for detection and model shapes for rendering. It can detecting intended line of users and draw exaggerated lines by changing the thickness of lines. In addition, we implemented a graphics API pipeline to run our method in real-time, and also examined a technique to accelerate processing speed.
