| OCR Text |
Show 24 within a single leaf-node bounding box) will not self-reflect or shadow (this is generally true if the bounding boxes are sufficiently dense). Thus if an intersection occurs within the same bounding box as the ray's origin, it can be rejected. However, this does not solve all of the problems. Consider the intersection depicted in Figure 2.6, where I is the incident ray, and S is the subsequent shadow ray cast to the light source. Because bounding box B2 is the closest to the origin of ray I, the Newton iteration starts there, and returns a successful intersection. It actually converges however, within bounding box Bl at point p2. Thus when S is cast, it registers an intersection at pl within bounding box Bl (since it is the closest). Since Bl and B2 are different, this could cause a self-intersection (i.e, the 'pointer test' described above would not detect a self-intersection). Currently, the solution to this problem is rather brute force: if an intersection converges to a point outside of a bounding box it was started in, it is rejected. While this does cause some extra intersection calculations to be performed, it is successful in avoiding the problem. To light Figure 2.6. An occurrence of self-intersection |
