Efficient Ray Tracing of Sparse Voxel Octrees on an FPGA

Ray tracing of sparse voxel octrees is a method of rendering images of 3D models, which could soon become practical for use in real time applications. This is desirable as ray tracing can produce very realistic visualizations, while voxel models can represent models with very fine geometric detail....

Full description

Bibliographic Details
Main Author: Wilhelmsen, Audun
Format: Others
Language:English
Published: Norges teknisk-naturvitenskapelige universitet, Institutt for elektronikk og telekommunikasjon 2012
Subjects:
Online Access:http://urn.kb.se/resolve?urn=urn:nbn:no:ntnu:diva-19415
Description
Summary:Ray tracing of sparse voxel octrees is a method of rendering images of 3D models, which could soon become practical for use in real time applications. This is desirable as ray tracing can produce very realistic visualizations, while voxel models can represent models with very fine geometric detail. For these reason the method has attracted significant attention in recent years, but no hardware solution has been published yet. This thesis presents a design of ray tracing of sparse voxel octrees in hardware. The objective is to show if it is sensible to implement the method in hardware, and if it could be integrated on modern GPUs alongside rasterization. To this end, the techniques used in existing software implementations of this method is reviewed, and an algorithm suitable for hardware implementation is presented. The problems of integrating the method with rasterization is explored, and the algorithm is analyzed and optimized to improve efficiency in hardware. A software implementation is presented, which supports the development of a hardware design. This design is implemented using the Verilog hardware description language, and it has been simulated and synthesized for an FPGA prototype. Multiple versions of the design has been synthesized and tested, and to evaluate the impact of design parameters the test results from these designs is presented. The thesis provides a comprehensive evaluation of the proposed design, and the results indicate that the algorithm is well suited for hardware implementation. Although real-time performance was not achieved, there are indications that further optimizations should allow real-time performance on the same platform, and that a full scale implementation on a modern GPU could probably allow ray tracing with a quality which is competitive with rasterization.