SpiralDiff：面向跨相机 RGB-to-RAW 转换的螺旋扩散与 LoRA
SpiralDiff: Spiral Diffusion with LoRA for RGB-to-RAW Conversion Across Cameras
把信号相关的噪声加权引入扩散采样过程，并按相机插入轻量 LoRA 分支，为「一个模型适配多台相机 ISP」提供了可复用的条件化与加权思路，可直接迁移到去噪、低光照增强等跨设备复原任务。
原文依据 a signal-dependent noise weighting strategy that adapts reconstruction fidelity across intensity levels
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论与 RGB 图像相比，RAW 图像保留了更优的保真度和更丰富的场景信息，因而在困难成像条件下的任务中不可或缺。为缓解数据采集的高昂成本，近期的 RGB-to-RAW 转换方法致力于从 RGB 图像合成 RAW 图像。然而，这些方法忽视了两个关键挑战：(i) 重建难度随像素强度而变化；(ii) 多相机转换需要针对具体相机的适配。为解决上述问题，我们提出 SpiralDiff，一个专为 RGB-to-RAW 转换设计的扩散框架，其信号相关的噪声加权策略可在不同强度水平上自适应地调整重建保真度。此外，我们引入 CamLoRA，一个相机感知的轻量适配模块，使统一模型能够适配不同相机特有的 ISP 特性。在四个基准数据集上的大量实验表明，SpiralDiff 在 RGB-to-RAW 转换质量上具有优越性，并在基于 RAW 的目标检测中带来下游收益。我们的代码已公开于 https://github.com/Chuancy-TJU/SpiralDiff。
英文原文摘要 1087 字符
RAW images preserve superior fidelity and rich scene information compared to RGB, making them essential for tasks in challenging imaging conditions. To alleviate the high cost of data collection, recent RGB-to-RAW conversion methods aim to synthesize RAW images from RGB. However, they overlook two key challenges: (i) the reconstruction difficulty varies with pixel intensity, and (ii) multi-camera conversion requires camera-specific adaptation. To address these issues, we propose SpiralDiff, a diffusion based framework tailored for RGB-to-RAW conversion with a signal-dependent noise weighting strategy that adapts reconstruction fidelity across intensity levels. In addition, we introduce CamLoRA, a camera-aware lightweight adaptation module that enables a unified model to adapt to different camera-specific ISP characteristics. Extensive experiments on four benchmark datasets demonstrate the superiority of SpiralDiff in RGB-to-RAW conversion quality and its downstream benefits in RAW-based object detection. Our code is available at https://github.com/Chuancy-TJU/SpiralDiff.