ColorFLUX: A Structure-Color Decoupling Framework for Old Photo Colorization

2026-03-30

AI Summary

This work addresses the challenge of colorizing aged photographs, which exhibit luminance decay and hue shifts that introduce significant domain discrepancies from modern images and hinder accurate color restoration by existing methods. To tackle this, the authors propose a structure–color disentangled framework based on the FLUX diffusion model, decoupling structural preservation from color recovery. The approach incorporates progressive Direct Preference Optimization (Pro-DPO) and a visual-semantic prompting mechanism to effectively mitigate text-prompt bias and domain shift. Evaluated on both synthetic and real-world datasets, the method outperforms current state-of-the-art techniques—including proprietary commercial models—producing vividly colored and structurally coherent results.

Abstract

Old photos preserve invaluable historical memories, making their restoration and colorization highly desirable. While existing restoration models can address some degradation issues like denoising and scratch removal, they often struggle with accurate colorization. This limitation arises from the unique degradation inherent in old photos, such as faded brightness and altered color hues, which are different from modern photo distributions, creating a substantial domain gap during colorization. In this paper, we propose a novel old photo colorization framework based on the generative diffusion model FLUX. Our approach introduces a structure-color decoupling strategy that separates structure preservation from color restoration, enabling accurate colorization of old photos while maintaining structural consistency. We further enhance the model with a progressive Direct Preference Optimization (Pro-DPO) strategy, which allows the model to learn subtle color preferences through coarse-to-fine transitions in color augmentation. Additionally, we address the limitations of text-based prompts by introducing visual semantic prompts, which extract fine-grained semantic information directly from old photos, helping to eliminate the color bias inherent in old photos. Experimental results on both synthetic and real datasets demonstrate that our approach outperforms existing state-of-the-art colorization methods, including closed-source commercial models, producing high-quality and vivid colorization.

Problem

• old photo colorization
• color degradation
• domain gap
• color bias
• historical photo restoration

Innovation

• structure-color decoupling
• generative diffusion model
• progressive Direct Preference Optimization
• visual semantic prompts
• old photo colorization