AI Paper Atlas

Research map / 3D vision / 3D generation

CHROMA: Consistent Harmonization of Multi-View Appearance via Bilateral Grid Prediction

Jisu Shin, Richard Shaw, Seunghyun Shin, Zhensong Zhang, Hae-Gon Jeon, Eduardo Pérez-Pellitero

ICLR 2026 · 3D vision › 3D generation · approach: Diffusion-based novel view synthesis

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
process the whole multi-view sequence jointly instead of single frames (ablation)CHROMA · LOMPSNR CC: 26.19 -> 26.25Improved
use proposed reference selection instead of a random reference frame (ablation)CHROMA · LOMPSNR CC: 25.11 -> 26.25Improved
combine exposure/entropy score with DINO instead of DINO-only reference selection (ablation)CHROMA · LOMPSNR CC: 25.95 -> 26.25Improved
add self-supervised rendering loss with unpaired WildRGBD data (ablation)CHROMA · BilaRFPSNR CC: 25.21 -> 25.60Improved
add self-supervised loss (ablation)CHROMA · BilaRFSSIM CC: 0.8245 -> 0.8240No clear effect
add CHROMA harmonization to DashGSDashGS · DL3DV w/ ISP variationPSNR: 23.35 -> 26.45Improved
add CHROMA harmonization to 3DGS3DGS · DL3DV w/ ISP variationPSNR: 21.43 -> 25.26Improved
add CHROMA harmonization to 3DGS3DGS · LOM varying exposurePSNR CC: 21.00 -> 26.25Improved
add CHROMA harmonization to DashGSDashGS · LOM varying exposurePSNR: 17.01 -> 18.19Mixed
add CHROMA harmonization to 3DGS3DGS · DL3DV w/ ISP variationtraining time: 3m 39s -> 3m 55sHurt
use feed-forward multi-view harmonization instead of per-frame 2D exposure correction (UEC)3DGS pipeline · LOM varying exposuretemporal flickering: 0.8124 -> 0.8428Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

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