The paper presents a deferred shading method to effectively render specular reflection with Gaussian splatting. The key challenge comes from the environment map reflection model, which requires accurate surface normal while simultaneously bottlenecks normal estimation with discontinuous gradients.
The authors leverage the per-pixel reflection gradients generated by deferred shading to bridge the optimization process of neighboring Gaussians, allowing nearly correct normal estimations to gradually propagate and eventually spread over all reflective objects. This enables high-quality rendering of specular reflection effects, demonstrating a consistent improvement of peak signal-to-noise ratio (PSNR) for both synthetic and real-world scenes, while running at a frame rate almost identical to vanilla Gaussian splatting.
The rendering pipeline consists of two passes:
The authors also propose a training algorithm featuring normal propagation, which expands reflective Gaussians to propagate their normal vectors to nearby Gaussians, enabling gradual propagation of correct normal estimation.
Experimental results show that the proposed method significantly outperforms state-of-the-art techniques and concurrent work in synthesizing high-quality specular reflection effects, while maintaining real-time performance. The method also produces more accurate normal and environment map estimation compared to other approaches.
Til et andet sprog
fra kildeindhold
arxiv.org
Vigtigste indsigter udtrukket fra
by Keyang Ye,Qi... kl. arxiv.org 04-30-2024
https://arxiv.org/pdf/2404.18454.pdfDybere Forespørgsler