About

References

Rendering point clouds with compute shaders
In some cases, compute shaders can render point clouds multiply times faster than native primitives like GL_POINTS or "point-list".
Fast Out-of-Core Octree Generation for Massive Point Clouds
Generating LOD structures can be slow and time consuming. Using hierarchical counting sort, we can efficiently organize the input into small batches that can be processed in parallel.
Potree: Rendering Large Point Clouds in Web Browsers
Rendering billions of points in web browsers using an octree acceleration structure.

QSplat
First approach to render large point clouds (or meshes transformed to point clouds) in real-time using a bounding-sphere hierarchy.
Layered Point Clouds
Basis of most commonly used point cloud acceleration structures, including Potree, Entwine, Arena4D, etc. The key feature is a GPU-friendly data structure that stores subsamples comprising thousands of points in each node.
High-Quality Surface Splatting on Today’s GPUs
Anti-Aliasing for point clouds by blending overlapping fragments within a certain range together. Results are similar to mip mapping or anisotropic filtering for textured meshes, which don't work for point clouds because they are colored by vertex rather than texture.
"A GPGPU-based Pipeline for Accelerated Rendering of Point Clouds."
Compute shader rendering using OpenCL.
"View-warped Multi-view Soft Shadowing for Local Area Lights"
Uses compute shaders for point-based depth map rendering.

This prototype of Potree 2 uses the AGPL license: https://www.gnu.org/licenses/agpl-3.0.en.html.