DiffHDR: Re-Exposing LDR Videos with Video Diffusion Models

2026-04-07

AI Summary

This work addresses the challenge of reconstructing high dynamic range (HDR) video from 8-bit low dynamic range (LDR) inputs, where highlight and shadow details are often lost due to saturation and quantization, hindering accurate radiance recovery. To tackle this, we introduce—for the first time—a video diffusion model for LDR-to-HDR reconstruction, formulating the task as generative radiance inpainting in a Log-Gamma color space latent representation. Our approach leverages spatiotemporal priors from a pretrained video diffusion model and supports controllable conversion guided by either text prompts or reference images. Furthermore, we devise a synthetic data generation strategy based on static HDRI maps to construct high-quality HDR video training pairs. Experiments demonstrate that our method significantly outperforms existing techniques in both radiometric fidelity and temporal consistency, producing photorealistic HDR videos capable of supporting substantial re-exposure adjustments.

Abstract

Most digital videos are stored in 8-bit low dynamic range (LDR) formats, where much of the original high dynamic range (HDR) scene radiance is lost due to saturation and quantization. This loss of highlight and shadow detail precludes mapping accurate luminance to HDR displays and limits meaningful re-exposure in post-production workflows. Although techniques have been proposed to convert LDR images to HDR through dynamic range expansion, they struggle to restore realistic detail in the over- and underexposed regions. To address this, we present DiffHDR, a framework that formulates LDR-to-HDR conversion as a generative radiance inpainting task within the latent space of a video diffusion model. By operating in Log-Gamma color space, DiffHDR leverages spatio-temporal generative priors from a pretrained video diffusion model to synthesize plausible HDR radiance in over- and underexposed regions while recovering the continuous scene radiance of the quantized pixels. Our framework further enables controllable LDR-to-HDR video conversion guided by text prompts or reference images. To address the scarcity of paired HDR video data, we develop a pipeline that synthesizes high-quality HDR video training data from static HDRI maps. Extensive experiments demonstrate that DiffHDR significantly outperforms state-of-the-art approaches in radiance fidelity and temporal stability, producing realistic HDR videos with considerable latitude for re-exposure.

Problem

• LDR-to-HDR conversion
• dynamic range expansion
• radiance recovery
• overexposed regions
• underexposed regions

Innovation

• video diffusion model
• LDR-to-HDR conversion
• radiance inpainting
• Log-Gamma color space
• controllable generation