Shadow Volume Reconstruction

generate shadows using several techniques, the most general being the shadow volume and the shadow map techniques. I will quickly survey these techniques and
a number of other techniques for casting shadows. Unfortunately, shadow volumes require access to a data structure representing the scene as a polygonal model,
and shadow maps currently require extensions to (and interfere with the use of) texture mapping hardware. While the restriction of shadow volumes to polygonal
databases does not sound severe, it does pose a modularity challenge in graphics APIs, and does not permit the use of shadow volumes with techniques such as
frame-buffer CSG. However, while shadow maps are not widely supported, the stencil planes required for shadow volumes are now available almost
universally. A hybrid of the shadow map and shadow volume algorithms will be presented which addresses these difficulties. The scene is rendered from the
point of view of the light source and a sampled depth map is recovered. Computer vision techniques (edge detection and vectorization) are used to reconstruct a
minimal shadow volume, after which the shadows can be rendered using only a one-bit stencil buffer. The shadow volume reconstruction algorithm also
addresses an important difficulty with the correct implementation of shadow volumes: capping the shadow volume boundary after the near and far-plane clips. I
will also demonstrate that the necessary computer vision algorithms can be implemented at real-time rates. I will also survey other research taking place within
CGL.