KLIP：逆问题中基于 KL 散度与扩散先验的局部分布偏移检测
KLIP: Localized Distribution Shift Detection via KL-Divergence with Diffusion Priors in Inverse Problems
把扩散模型当作数据先验，用先验与后验之间的 KL 散度作为无需校准数据的 OOD 指标，并可在图像块级别定位分布偏移；对做逆问题求解（如 CT 重建）的视觉研究者，这提供了一种在只有间接测量时判断输入是否可信、并定位异常区域的通用手段。
原文依据 an OOD detection metric based on the Kullback-Leibler divergence between the diffusion prior and the posterior distribution
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论扩散模型作为计算成像的数据驱动先验已展现出良好性能，同时也具备一定的分布外（OOD）图像检测能力。然而，现有的 OOD 检测方法往往需要关于偏移分布的部分知识，无法检测细微或局部的分布偏移，并且作用于完整图像，而非逆问题中可获得的间接测量。我们提出一种基于扩散先验与后验分布之间 Kullback-Leibler 散度的 OOD 检测指标，该指标（i）不需要任何校准数据或关于偏移分布的知识，（ii）既可将整幅图像判定为 OOD，也可在图像内定位 OOD 图像块。实验表明，该指标能够检测细微但具有语义意义的分布偏移，例如从健康肝脏 CT 扫描到带肿瘤扫描的偏移，并可泛化到不同类型的扩散模型、数据集与逆问题。
英文原文摘要 1006 字符
Diffusion models have shown promising performance as data-driven priors for computational imaging, as well as some capacity to detect out-of-distribution (OOD) images. However, existing approaches to OOD detection often require some knowledge of the shifted distribution, fail to detect subtle or localized distribution shifts, and operate on full images, rather than the indirect measurements available in inverse problems. We propose an OOD detection metric based on the Kullback-Leibler divergence between the diffusion prior and the posterior distribution, that (i) does not require any calibration data or knowledge of the shifted distribution, and (ii) can detect whole images as OOD as well as localize OOD patches within an image. Experimentally, we show that this metric can detect subtle yet semantically meaningful distribution shifts, such as the shift from healthy liver CT scans to those with tumors, and generalizes across different types of diffusion models, datasets, and inverse problems.