What question did this study set out to answer?

This research aims to create a rendering pipeline that accurately models defocus effects in 3D Gaussian Splatting.

Closed-Form Convolution for physically-accurate defocus in Gaussian Splatting

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial reveals improved defocus rendering in multi-focus images, suggesting enhanced artistic control.

This research aims to create a rendering pipeline that accurately models defocus effects in 3D Gaussian Splatting.
Developed a physically accurate depth-of-field model integrated into the 3D Gaussian Splatting rasterization process.
Derived a closed-form analytical solution for convolving each projected 3D Gaussian with a kernel based on Circle of Confusion.
Extended methodology to handle anisotropic Gaussians and approximate polygonal apertures.
The new method produced sharper in-focus details and smoother transitions in defocus compared to existing techniques.
Achieved competitive performance on defocus benchmarks with improved visual fidelity.
Enabled rich post-capture effects like refocusing and aperture manipulation in 3D reconstructions.

Dai et al. (2026) studied this question.

View Full Paper

Ask AI

Bookmark