TRIPS: Trilinear Point Splatting for Real-Time Radiance Field Rendering

TRIPS: Trilinear Point Splatting for Real-Time Radiance Field Rendering

26 Mar 2024 | Linus Franke, Darius Rückert, Laura Fink and Marc Stamminger
TRIPS (Trilinear Point Splatting) is a novel approach for real-time radiance field rendering, combining the strengths of both 3D Gaussian Splatting and ADOP (Approximate Differentiable One-Pixel Point Rendering). The method rasterizes points into a screen-space image pyramid, allowing for the rendering of large points using trilinear write operations. A lightweight neural network reconstructs hole-free images, enhancing texture details beyond the resolution of the original point cloud. The pipeline is differentiable, enabling automatic optimization of point sizes and positions. Evaluations demonstrate that TRIPS outperforms existing methods in terms of rendering quality while maintaining a real-time frame rate of 60 frames per second on standard hardware. The approach is robust to challenging scenarios, including intricate geometry, expansive landscapes, and auto-exposed footage. The project page and implementation details are available at <https://lfranke.github.io/trips>.TRIPS (Trilinear Point Splatting) is a novel approach for real-time radiance field rendering, combining the strengths of both 3D Gaussian Splatting and ADOP (Approximate Differentiable One-Pixel Point Rendering). The method rasterizes points into a screen-space image pyramid, allowing for the rendering of large points using trilinear write operations. A lightweight neural network reconstructs hole-free images, enhancing texture details beyond the resolution of the original point cloud. The pipeline is differentiable, enabling automatic optimization of point sizes and positions. Evaluations demonstrate that TRIPS outperforms existing methods in terms of rendering quality while maintaining a real-time frame rate of 60 frames per second on standard hardware. The approach is robust to challenging scenarios, including intricate geometry, expansive landscapes, and auto-exposed footage. The project page and implementation details are available at <https://lfranke.github.io/trips>.
Reach us at info@study.space