Naka-GS: Naka-Guided Chroma Correction and Point-Prior Compaction for Low-Light 3D Gaussian Splatting

NAKA-GS improves 3D Gaussian Splatting for low-light scenes by combining photometric enhancement and geometric refinement. It uses a Naka-inspired algorithm to enhance visibility and then employs a learnable network to correct color errors while preserving fine details. A point cloud preprocessing module further cleans the geometry before training, resulting in better reconstruction quality.

Open the full episode

More from AI Radio