Title

GPU-ACCELERATED PRECOMPUTED DISCRETE VISIBILITY FIELDS FOR REAL-TIME RAY-TRACED DYNAMIC SCENES UNDER ENVIRONMENT LIGHTING

Abstract

Abstract

Environment lighting plays a crucial role in enhancing the realism of rendered images.
Environment light sources are distributed over a sphere or hemisphere, and the
rendering process consists of tracing multiple rays per pixel. This process is computationally
expensive, therefore achieving real-time results is challenging. Many studies
focus on visibility function to reduce the number of traced rays, often achieved
through precomputation methods.
This study improves upon a CPU-based visibility precomputation method designed
for rendering solely static scenes. CPU-based precomputation method is presented
in detail, it is implemented, and illustrations for each step are provided. Then, a
GPU-based precomputation solution that supports rendering of dynamic scenes is
introduced. The solution is implemented with multiple compute shaders. To demonstrate
the capability of dynamic scene rendering, a proof-of-concept CPU animation
technique is employed.
Static scene rendering performance is evaluated by replicating the tests performed
in the CPU-based precomputation method, and improvements are demonstrated with
higher rendering speed and more accurate results. Furthermore, additional tests are
conducted to assess the dynamic scene rendering performance. Successful results
suggest promising avenues for future studies involving GPU animation.

Supervisor(s)

Supervisor(s)

BERIL GUNAY

Date and Location

Date and Location

2024-01-18 12:30:00

Category

Category

MSc_Thesis