GMODiff: One-Step Gain Map Refinement with Diffusion Priors for HDR Reconstruction

2025-12-18

AI Summary

To address three key challenges in multi-exposure HDR reconstruction using pretrained latent diffusion models (LDMs)—limited dynamic range, high inference overhead, and content hallucination—this paper proposes a gain-map (GM)-driven single-step diffusion framework. Instead of generating full HDR images, our method focuses on estimating the GM as the primary task, leveraging LDMs’ latent-space priors for efficient, single-step reconstruction. We introduce a novel integration of regression-guided denoising and latent-code decoding, jointly optimizing structural fidelity and perceptual quality while effectively suppressing hallucinations. Experiments demonstrate that our approach achieves state-of-the-art performance across multiple HDR reconstruction metrics. Moreover, it accelerates inference by 100× compared to existing LDM-based methods, significantly reducing computational cost without sacrificing quality.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Learning to Search

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Pre-trained Latent Diffusion Models (LDMs) have recently shown strong perceptual priors for low-level vision tasks, making them a promising direction for multi-exposure High Dynamic Range (HDR) reconstruction. However, directly applying LDMs to HDR remains challenging due to: (1) limited dynamic-range representation caused by 8-bit latent compression, (2) high inference cost from multi-step denoising, and (3) content hallucination inherent to generative nature. To address these challenges, we introduce GMODiff, a gain map-driven one-step diffusion framework for multi-exposure HDR reconstruction. Instead of reconstructing full HDR content, we reformulate HDR reconstruction as a conditionally guided Gain Map (GM) estimation task, where the GM encodes the extended dynamic range while retaining the same bit depth as LDR images. We initialize the denoising process from an informative regression-based estimate rather than pure noise, enabling the model to generate high-quality GMs in a single denoising step. Furthermore, recognizing that regression-based models excel in content fidelity while LDMs favor perceptual quality, we leverage regression priors to guide both the denoising process and latent decoding of the LDM, suppressing hallucinations while preserving structural accuracy. Extensive experiments demonstrate that our GMODiff performs favorably against several state-of-the-art methods and is 100 faster than previous LDM-based methods.

Problem

• Overcoming limitations of Latent Diffusion Models for HDR reconstruction.
• Reducing high inference cost from multi-step denoising in HDR tasks.
• Mitigating content hallucination while preserving structural accuracy in HDR.

Innovation

• One-step diffusion framework using gain maps
• Regression-based initialization for single denoising step
• Guided denoising with regression priors to reduce hallucinations