Hybrid Scene Structuring: Application to Ray Tracing
150 likes | 260 Views
Explore a new algorithm for efficient ray acceleration, featuring bounding volume optimization, node classification, and uniform space subdivision. Enhance your ray tracing speed and space utilization with this advanced technique.
Hybrid Scene Structuring: Application to Ray Tracing
E N D
Presentation Transcript
Hybrid Scene Structuringwith Application to Ray Tracing Gordon Müller, Dieter Fellner Institute of ComputerGraphics, TU Braunschweig {gordon.mueller,d.fellner}@tu-bs.de Gordon Müller, Dieter Fellner
Overview • Ray Acceleration • Our Algorithm • Bounding Volume Optimization • Node Classification • Space Subdivision • Results • Conclusions Gordon Müller, Dieter Fellner
5 6 1 3 7 4 2 8 5 6 1 3 7 4 2 8 Ray Accelerationprevious work • Hierarchical bounding volumes(Kay/Kajiya 1986, Goldsmith/Salmon 1987) • Uniform space subdivision(Fujimoto et al. 1986) • Octrees, BSP-trees, …(Glassner 1984,…) • Adaptive grids, HUG(Cazals et al. 1995) Gordon Müller, Dieter Fellner
Algorithm Overview • Hierarchical bounding volume construction based on a cost function • Identification of uniform regions by classification of sub-scenes • Locally space subdivision of uniform regions Gordon Müller, Dieter Fellner
Step 1:Bounding Volume Optimization • Recursively subdivide the set of objects into two disjoint sub-scenes • Objects are sorted along coordinate axes • No fixed subdivision position... Gordon Müller, Dieter Fellner
Step 1:Bounding Volume Optimization • …instead, we minimize a cost function describing the approximated ray/scene intersection costs • Object-specific intersection costs • O(n logn) on average Gordon Müller, Dieter Fellner
Example (subdivision) Gordon Müller, Dieter Fellner
Step 2: Node Classification • Goal: detect scene nodes that are suitable base nodes for a uniform space subdivision • Recursively classify scene nodes based on • surface area of neighbor hierarchy nodes • volume of neighbor hierarchy nodes • average size of elementary objects below a hierarchy node • Threshold constants determined empirically Gordon Müller, Dieter Fellner
Step 2: Node Classification • Scene nodes hold a counter representing the number of uniform classified sub-nodes(used in step 3) • Classification does not destroy hierarchy! • O(n) Gordon Müller, Dieter Fellner
Example (classification) Gordon Müller, Dieter Fellner
Step 3: Uniform Space Subdivision • Build uniform space subdivisions for sub-scenes marked in the previous step • Recursively subdivide the bounding box of a scene node along the dominant axis • Sub-node counter used to determine number of voxels / subdivisions • Use available bounding volume hierarchy to speed-up voxel membership tests ! • O(n) on average Gordon Müller, Dieter Fellner
Timings Gordon Müller, Dieter Fellner
Conclusions • Hybrid ray acceleration • run-time efficient • space efficient • easy to implement • easy to use Gordon Müller, Dieter Fellner
Future Work • Bounding volume hierarchies • View frustum/occlusion culling • Dynamic environments • Collision detection • Parallelization • Clustering • Hierarchical Radiosity Gordon Müller, Dieter Fellner
Thank you! Gordon Müller, Dieter Fellner