Parallel BVH Construction Using Locally Density Clustering

Yingsong Hu,Qingzhi Zeng,Dan Li,Weijian Wang,Yunfei Hu

doi:10.1109/access.2019.2932151

Abstract

A novel bounding volume hierarchy (BVH) construction method based on locally dense clustering (LDC) was proposed for the low quality of BVH constructed in a complex scene with uneven distribution of primitives. The quality of the BVH was effectively improved by defining primitive density, analyzing the relation between density and traversal efficiency, selecting primitives with high density as the clustering center at the early stages of construction, and combining with top-down iterative clustering and bottom-up agglomerative clustering (AC). In order to effectively calculate primitive density, a local search strategy based on Morton coding was adopted. This strategy can quickly get approximate primitive density information through GPU. We evaluate the method and show that, in the complex building scene with uneven distribution of primitives, our method is 13% higher in traversal speed and 11% lower in surface area heuristic (SAH) cost. That means our method can improve the rendering speed of ray tracing effectively.

Highlights

Ray tracing is a global illumination rendering algorithm
Since the surface area heuristic (SAH) cost function can only express the approximate expected cost of the whole acceleration structure of ray traversal and the ray may exit the traversal process very early in the actual traversal process, the SAH cost cannot be used as single evaluation criteria
The time of bounding volume hierarchy (BVH) construction, SAH cost, traversal speed, overlap and other metrics in this test are recorded for detailed analysis

Summary

INTRODUCTION

Ray tracing is a global illumination rendering algorithm. Its physical principle based on ray propagation can better render the light phenomena of the scene such as blurring, shading, and highlighting, making the image more realistic and closer to the goal of photo-quality image pursued by humans. In the actual calculation process, billions of rays may be required for the intersection test with millions of primitives in the scene. Weller et al [2] proposed to construct BVH on the GPU by the clustering algorithm Batch Neural Gas (BNG) of machine learning. Their method is not for the 3D scenes represented by the mesh model, but for those represented by the voxel. The construction of BVH is to divide the graphics primitives in 3D scenes and is a hierarchical clustering problem. We proposed, by analyzing the distribution of primitive density, to construct BVH on the GPU by the method based on LDC. In most scenes with uneven distribution of primitives, our SAH cost, traversal speed, end-point overlap (EPO) [4] and other quantitative criteria are superior to their method

RELATED WORK

DEFINITION OF PRIMITIVE DENSITY

BVH CONSTRUCTION VIA LDC

Result

RESULTS AND ANALYSIS

CONCLUSION