通过捷径采样加速扩散模型求解逆问题
Accelerating Diffusion Models for Inverse Problems through Shortcut Sampling
AI总结:
本文提出扩散捷径采样(SSD),通过寻找连接测量图像与恢复图像的过渡状态并引入失真自适应反演和反投影约束,在仅30 NFE下实现与100 NFE零样本方法相当甚至更优的逆问题求解效果。
摘要:
扩散模型近期展现出以无监督方式解决逆问题的出色能力。现有方法主要聚焦于修改后验采样过程,而前向过程的潜力在很大程度上仍未得到探索。本文提出一种用于零样本求解逆问题的全新方法——扩散捷径采样(Shortcut Sampling for Diffusion,SSD)。SSD的核心思想并非从随机噪声出发,而是寻找一个连接测量图像y与恢复图像x的特定过渡状态。通过利用“输入—过渡状态—输出”的捷径路径,SSD能够以更少的步骤实现精确恢复。为了在前向过程中推导过渡状态,我们引入了失真自适应反演(Distortion Adaptive Inversion)。此外,在生成过程中我们应用反投影作为额外的一致性约束。实验表明,SSD在多个代表性图像恢复(IR)任务上具有有效性。与最先进的零样本方法(100 NFE)相比,我们的方法仅用30 NFE即可取得有竞争力的结果,并在某些任务中以100 NFE超越这些方法。代码见 https://github.com/GongyeLiu/SSD。
英文摘要:
Diffusion models have recently demonstrated an impressive ability to address inverse problems in an unsupervised manner. While existing methods primarily focus on modifying the posterior sampling process, the potential of the forward process remains largely unexplored. In this work, we propose Shortcut Sampling for Diffusion(SSD), a novel approach for solving inverse problems in a zero-shot manner. Instead of initiating from random noise, the core concept of SSD is to find a specific transitional state that bridges the measurement image y and the restored image x. By utilizing the shortcut path of "input - transitional state - output", SSD can achieve precise restoration with fewer steps. To derive the transitional state during the forward process, we introduce Distortion Adaptive Inversion. Moreover, we apply back projection as additional consistency constraints during the generation process. Experimentally, we demonstrate SSD's effectiveness on multiple representative IR tasks. Our method achieves competitive results with only 30 NFEs compared to state-of-the-art zero-shot methods(100 NFEs) and outperforms them with 100 NFEs in certain tasks. Code is available at https://github.com/GongyeLiu/SSD
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。