Score-Based Diffusion Models as Principled Priors for Inverse Imaging
Score-Based Diffusion Models as Principled Priors for Inverse Imaging
复制标题
DOI：
10.1109/iccv51070.2023.00965
复制
发表时间：
2023-04
期刊：
影响因子：
--
通讯作者：
Berthy T. Feng;Jamie Smith;Michael Rubinstein;Huiwen Chang;K. Bouman;W. T. Freeman
中科院分区：
文献类型：
--
作者：
Berthy T. Feng;Jamie Smith;Michael Rubinstein;Huiwen Chang;K. Bouman;W. T. Freeman
先验对于从噪声和/或不完整的测量重建图像是必不可少的。先验的选择决定了恢复图像的质量和不确定性。我们建议将基于分数的扩散模型转化为原则性的图像先验（“基于分数的先验”），用于分析给定测量的图像的后验。以前，概率先验仅限于手工正则化器和简单分布。在这项工作中，我们经验验证了理论证明的概率函数的分数为基础的扩散模型。我们展示了如何使用变分推理的概率函数从所得后验样本。我们的研究结果，包括去噪，去模糊和干涉成像的实验，表明基于分数的先验知识，使一个复杂的，数据驱动的图像先验的原则性推理。
Priors are essential for reconstructing images from noisy and/or incomplete measurements. The choice of the prior determines both the quality and uncertainty of recovered images. We propose turning score-based diffusion models into principled image priors ("score-based priors") for analyzing a posterior of images given measurements. Previously, probabilistic priors were limited to handcrafted regularizers and simple distributions. In this work, we empirically validate the theoretically-proven probability function of a score-based diffusion model. We show how to sample from resulting posteriors by using this probability function for variational inference. Our results, including experiments on denoising, deblurring, and interferometric imaging, suggest that score-based priors enable principled inference with a sophisticated, data-driven image prior.