| OCR Text |
Show CHAPTER 5 OCTREE RAY INTERSECTION METHODS SOME EMPIRICAL RESULTS Since ray tracing was used so heavily in the shading model described, some investigation of efficient methods for ray intersections was performed. An existing ray tracer which used hierarchical bounding volumes simjlar to the method described by Rubin and Whitted (30,27] was modified to use octree data structures. This chapter includes results from test runs with the octree implementation. Octree performance with different octree parameters is tested, along with a brief analysis. Timjngs of three different ray intersection methods are included in Chapter 6. 5.1 Octree Implementation The octree implementation followed Glassner's technique closely. Nodes of the octree were stored in a hash table, where the label and depth of. a node were used to hash into a. bucket. {The label of a node is an encoding of the path down the · octree to the node.) After hashing into a bucket, a linear search was used to find a matching key. Since a linear search wa.s used in the buckets, the number of buckets in the hash table was relatively large to avoid high search costs. A significant portion of the execution time of this ray intersection method was spent in point classification in the octree, and the hash table was usually accessed several times to classify a point. The hash table was important to overall performance, and was modified dynamically to adapt to queries and improve its speed. |
