Abstract:
|
Silhouette detection is a key issue in many non-photorealistic rendering algorithms. Silhouettes play an important role in shape recognition because they provide one of the main cues for figure-to-ground distinction. However, since silhouettes are view-dependent, they must be computed per frame. There has been a lot of research tailored
to find an efficient way to do this in real time, although most algorithms require either a precomputation, and/or multiple rendering passes. In this paper we present an algorithm that is able to generate silhouettes in realtime by using the geometric shader stage. In contrast to other approaches, our silhouettes do not suffer from discontinuities. A second contribution is the definition of a continuous set of texture coordinates that allows us coherently texture the object, independently on the length and orientation of the edges. This texture coordinates are also robust to translation and does not produce sudden changes during rotation. |