Skip to main content
QUICK REVIEW

[论文解读] HDR Reconstruction Boosting with Training-Free and Exposure-Consistent Diffusion

Yo-Tin Lin, Su-Kai Chen|arXiv (Cornell University)|Feb 23, 2026
Image Enhancement Techniques被引用 0
一句话总结

一个无需训练的扩散式去塌陷框架通过合理填充过曝区域并在多曝光之间确保亮度与纹理的一致性来提升现有 HDR 重建方法,无需额外训练。

ABSTRACT

Single LDR to HDR reconstruction remains challenging for over-exposed regions where traditional methods often fail due to complete information loss. We present a training-free approach that enhances existing indirect and direct HDR reconstruction methods through diffusion-based inpainting. Our method combines text-guided diffusion models with SDEdit refinement to generate plausible content in over-exposed areas while maintaining consistency across multi-exposure LDR images. Unlike previous approaches requiring extensive training, our method seamlessly integrates with existing HDR reconstruction techniques through an iterative compensation mechanism that ensures luminance coherence across multiple exposures. We demonstrate significant improvements in both perceptual quality and quantitative metrics on standard HDR datasets and in-the-wild captures. Results show that our method effectively recovers natural details in challenging scenarios while preserving the advantages of existing HDR reconstruction pipelines. Project page: https://github.com/EusdenLin/HDR-Reconstruction-Boosting

研究动机与目标

  • 在信息丢失的严重过曝场景下,推动鲁棒的 HDR 重建。
  • 开发一个无需训练的流水线,通过扩散先验来增强间接和直接 HDR 方法。
  • 通过迭代的去噪和补偿策略确保跨曝光的亮度一致性。
  • 在不需要模型训练的前提下,将扩散式去噪与现有 HDR 流程整合。

提出的方法

  • 利用通过 ControlNet 条件化的深度/结构的扩散式去模版骨干来填充由文本提示引导的过曝区域。
  • 应用带有计划强度的迭代 SDEdit 精化,以在创造性和跨 EV 的一致性之间取得平衡。
  • 在每次去模版后执行补偿步骤,以强化亮度下界约束和跨 EV 的一致性。
  • 通过 Debevec 的方法对 LDR 堆栈进行 EV-bracket 图像生成和反 CRF 估计,以实现可靠的色调映射与 HDR 合并。
  • 实现覆盖掩膜和基于亮度的残差补偿,以逐步使生成内容与物理曝光关系对齐。
Figure 2 : Overview of our training-free HDR reconstruction pipeline. Given an input LDR image (EV0), we generate bracketed LDR images using an existing HDR reconstruction method. Our iterative pipeline then enhances these results through (1) an inpainting stage guided by exposure and condition maps
Figure 2 : Overview of our training-free HDR reconstruction pipeline. Given an input LDR image (EV0), we generate bracketed LDR images using an existing HDR reconstruction method. Our iterative pipeline then enhances these results through (1) an inpainting stage guided by exposure and condition maps

实验结果

研究问题

  • RQ1一个无需训练的扩散式去模版流水线是否能够在多基线方法的 HDR 重建中改进过曝区域的重建?
  • RQ2在不额外训练的情况下,如何在迭代去模版和 HDR 合并过程中保持跨曝光(EV)的一致性?
  • RQ3在迭代过程中,补偿在强制亮度下界和 CRF 稳定性方面起到的作用是什么?
  • RQ4所提出的方法如何在不同的扩散骨干和基线 HDR 技术之间实现泛化?

主要发现

  • 所提出的方法在提升基线 HDR 方法时,on VDS 和 HDR-Eye 数据集上对非参考图像质量指标表现出一致改进。
  • 该方法在过曝区域内提升感知真实感(例如天空细节),同时保持跨曝光的结构一致性。
  • 在无训练的前提下,结合扩散先验的框架在与 CEVR、SingleHDR、GlowGAN、Deep Recursive HDRI、Multi-Exposure Generation 等方法集成时,优于若干基线方法。
  • SDEdit 强度计划和补偿流程是维持跨 EV 连贯性、防止伪影的重要消融项。
  • 扩散骨干的可变性(SDXL 与 SDXL Turbo)在不重新训练的情况下仍然带来改进,展示了对骨干的鲁棒性。
Figure 3 : Limitations arising from naively combining indirect HDR reconstruction methods and over-exposed regions inpainting. Independent inpainting of each EV bracket, without cross-EV alignment, can introduce ghosting artifacts in the merged HDR result. These artifacts stem from inconsistencies i
Figure 3 : Limitations arising from naively combining indirect HDR reconstruction methods and over-exposed regions inpainting. Independent inpainting of each EV bracket, without cross-EV alignment, can introduce ghosting artifacts in the merged HDR result. These artifacts stem from inconsistencies i

更好的研究,从现在开始

从论文设计到论文写作,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。