AI Paper Atlas

Research map / 3D vision / Gaussian splatting

GaussianImage: 1000 FPS Image Representation and Compression by 2D Gaussian Splatting

Xinjie Zhang, Xingtong Ge, Tongda Xu, Dailan He, Yan Wang, Hongwei Qin, Guo Lu, Jing Geng, Jun Zhang

ECCV 2024 · 3D vision › Gaussian splatting · approach: Regularization and primitive design for 3DGS · 74 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
replace alpha blending with accumulated summation blending (ablation)GaussianImage · Kodak (30000 Gaussians)PSNR: 37.89 -> 38.69Improved
use 2D Gaussians instead of 3D Gaussians (ablation)GS image representation · Kodak (30000 Gaussians)PSNR / parameters: 37.41 -> 37.89 PSNR; 1770K -> 270K paramsImproved
merge color coefficients and opacity into weighted color (ablation)GaussianImage · Kodak (30000 Gaussians)PSNR: 38.69 -> 38.57; params 270K -> 240KNo clear effect
use L2 loss instead of L1, SSIM, or combinations (ablation)GaussianImage · KodakPSNR: 36.57 (L1+SSIM) -> 38.57Improved
Cholesky versus rotation-scaling covariance factorization (ablation)GaussianImage · KodakPSNR: 38.57 vs 38.83No clear effect
add commitment loss for RVQ codebook (ablation)GaussianImage codec · KodakBD-rate (%): 333.16 -> 0Improved
use RVQ for color instead of 6-bit integer quantization (ablation)GaussianImage codec · KodakBD-rate (%): 7.02 -> 0Improved
use higher-bit (8-bit) quantization for covariance/color (ablation)GaussianImage codec · KodakBD-rate (%): 7.02 -> 11.69Hurt
use 2D Gaussians versus I-NGP and NeuRBF for representationimage representation · Kodak / DIV2KGPU memory and training time: Kodak training 490.61 -> 106.59 s; memory 1525 -> 419 MiBImproved
add partial bits-back coding to codec (ablation)GaussianImage codec · Kodak/DIV2K rate-distortionrate-distortionImproved

Later papers that compared with this method

8 later accepted papers report GaussianImage as a baseline in their results tables; 8 of them report a better number on most of the results they share with it.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy