EVER: Exact Volumetric Ellipsoid Rendering for Real-time View Synthesis
Alexander Mai, Peter Hedman, George Kopanas, Dor Verbin, David Futschik, Qiangeng Xu, Falko Kuester, Jon Barron, Yinda Zhang
2024-10-03
Summary
This paper presents EVER, a new method for rendering 3D ellipsoids in real-time, which improves the quality of generated images without the common issues found in previous methods.
What's the problem?
Many current methods for rendering 3D images, like 3D Gaussian Splatting (3DGS), can produce visual artifacts such as 'popping' effects and inconsistent image quality when viewed from different angles. These problems occur because they often rely on simpler techniques that don't accurately represent the volume of objects, leading to poor image quality and performance issues.
What's the solution?
The authors introduce Exact Volumetric Ellipsoid Rendering (EVER), which uses a more precise method to render 3D shapes called ellipsoids. This approach allows for exact volume rendering instead of just layering images on top of each other. By using ray tracing, EVER can create high-quality effects like blur and distortion that are hard to achieve with older methods. The results show that EVER performs better than 3DGS, especially in complex scenes, while maintaining fast frame rates (around 30 frames per second) even at high resolutions.
Why it matters?
This research is important because it advances the field of real-time graphics rendering, making it possible to create more realistic and visually appealing images in applications like video games, simulations, and virtual reality. By addressing the limitations of previous methods, EVER can enhance user experiences in various digital environments.
Abstract
We present Exact Volumetric Ellipsoid Rendering (EVER), a method for real-time differentiable emission-only volume rendering. Unlike recent rasterization based approach by 3D Gaussian Splatting (3DGS), our primitive based representation allows for exact volume rendering, rather than alpha compositing 3D Gaussian billboards. As such, unlike 3DGS our formulation does not suffer from popping artifacts and view dependent density, but still achieves frame rates of sim!30 FPS at 720p on an NVIDIA RTX4090. Since our approach is built upon ray tracing it enables effects such as defocus blur and camera distortion (e.g. such as from fisheye cameras), which are difficult to achieve by rasterization. We show that our method is more accurate with fewer blending issues than 3DGS and follow-up work on view-consistent rendering, especially on the challenging large-scale scenes from the Zip-NeRF dataset where it achieves sharpest results among real-time techniques.