CanonCGT: Reference-Based Color Grading via Canonical Pivot Representation

2026-05-31

AI Summary

Existing reference-based color grading methods often produce unnatural results due to unstable tone mapping, leading to color shifts or inconsistencies. To address this issue, this work proposes CanonCGT, a two-stage framework that introduces a style-agnostic “canonical pivot” as an intermediate representation: the first stage removes inherent tonal biases from the input image, and the second stage accurately aligns it with the target reference style. The approach integrates a novel canonical pivot representation with a dual-phase training mechanism (DP-CGT), combining supervised pretext learning and self-supervised optimization on unpaired images. Experimental results demonstrate that CanonCGT significantly outperforms state-of-the-art methods across multiple datasets, achieving superior performance in photorealism, tonal consistency, and visual fidelity.

Abstract

Reference-based color grading aims to reproduce the tonal mood and lighting of a reference while preserving color harmony and scene structure. Existing photorealistic and filter-based methods often produce unstable tone mappings -- over-shifting or inconsistently retaining colors -- leading to unnatural results. We propose CanonCGT, a two-stage framework built on a canonical pivot -- a style-neutral intermediate representation for stable color mapping. The first stage canonicalizes the input by removing intrinsic tonal bias, and the second color-grades it to match the reference style. A dual-phase training scheme, DP-CGT, combines supervised preset learning with self-supervised refinement on unpaired photographs. CanonCGT delivers photorealistic and tonally consistent results across diverse datasets, surpassing state-of-the-art methods in stability and visual fidelity. Our codes are available at \href{https://github.com/Jinwon-Ko/CanonCGT}{https://github.com/Jinwon-Ko/CanonCGT}

Problem

• reference-based color grading
• tone mapping
• color consistency
• photorealistic rendering
• color harmony

Innovation

• canonical pivot
• reference-based color grading
• two-stage framework
• dual-phase training
• photorealistic tone mapping