| OCR Text |
Show 23 3.6 Summary The ray intersection methods described convert the objects in a scene into a data structure which reduces the size of a candidate set of objects to be tested for ray intersection. These methods must generate candidate sets guaranteeing any appropriate object is tested and the object's ray intersection will be found. Warren [35], Goldsm1th [17], and Arvo [3] provide some discussion of expected execution behavior. While it would be desirable to do a formal analysis of the complexity of ray intersection methods, there are several reasons why this would be difficult. Methods often use mixed algorithms - for example an octree based on a hash table, or hierarchical bounding volumes with a heap sort of a candidate set. Different systems use different geometric primitives, changing some of the computational cost relationships between data structure traversal and actual geometric primitive intersection tests. At this point, empirical comparisons of these techniques are published more often than theoretical analyses, providing a starting point for further work. |
