CARD：基于扩散的相关噪声感知复原
CARD: Correlation Aware Restoration with Diffusion
针对真实传感器中空间相关噪声的扩散复原方法，通过噪声白化将相关噪声转为独立同分布形式，并修改DDRM的采样更新步骤，为处理非i.i.d.噪声的复原任务提供了免训练、闭式采样的通用思路。
原文依据 CARD first whitens the noisy observation, converting the noise into an i.i.d. form.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论去噪扩散模型通过将图像复原建模为顺序去噪步骤，已在该领域取得最优性能。然而，大多数方法假设噪声为独立同分布的高斯噪声，而真实传感器由于读出机制往往表现出空间相关噪声，这在实际中限制了复原质量。我们提出相关噪声感知扩散复原（CARD），这是DDRM的一种免训练扩展，能够显式处理相关高斯噪声。CARD首先对含噪观测进行白化，将噪声转换为独立同分布形式。随后，扩散复原步骤被替换为噪声白化更新，这些更新继承了DDRM的闭式采样效率，同时支持相关噪声下的复原。为强调处理相关噪声的重要性，我们贡献了CIN-D，这是一个在不同光照条件下采集的新型相关噪声数据集，用于评估复原方法在真实卷帘快门传感器噪声上的表现。该数据集填补了文献中利用真实世界相关噪声进行实验评估的关键空白。在合成相关噪声的标准基准和CIN-D上的实验表明，CARD在去噪、去模糊和超分辨率任务中始终优于现有方法。
英文原文摘要 1343 字符
Denoising diffusion models have achieved state-of-the-art performance in image restoration by modeling the process as sequential denoising steps. However, most approaches assume independent and identically distributed (i.i.d.) Gaussian noise, while real-world sensors often exhibit spatially correlated noise due to readout mechanisms, limiting restoration quality in practice. We introduce Correlation Aware Restoration with Diffusion (CARD), a training-free extension of DDRM that explicitly handles correlated Gaussian noise. CARD first whitens the noisy observation, converting the noise into an i.i.d. form. The diffusion restoration steps are then replaced with noise-whitened updates, which inherit DDRM's closed-form sampling efficiency while enabling restoration under correlated noise. To emphasize the importance of addressing correlated noise, we contribute CIN-D, a novel correlated noise dataset captured across diverse illumination conditions to evaluate restoration methods on real rolling-shutter sensor noise. This dataset fills a critical gap in the literature for experimental evaluation with real-world correlated noise. Experiments on standard benchmarks with synthetic correlated noise and on CIN-D demonstrate that CARD consistently outperforms existing methods across denoising, deblurring, and super-resolution tasks.