PCDM: PERCEPTUAL CONSISTENCY IN DIFFUSION MODELS FOR NO-REFERENCE IMAGE QUALITY ASSESSMENT

Shreshth Saini, Ru-Ling Liao, Yan Ye, Alan Bovik

Keywords: No Reference Image Quality Assessment, Perceptual Image Quality, Latent Diffusion Models, Training-free

TL;DR: We propose Perceptual Manifold Guidance in Latent Diffusion Models to extract multi-scale, multi-time denoising U-Net features for zero-shot No-Reference Image Quality Assessment, marking the first-ever application of pretrained LDMs to NR-IQA.

Abstract: Despite recent advancements in latent diffusion models for generating high-dimensional data and performing various downstream tasks, there has been little exploration into perceptual consistency within these models for No-Reference Image Quality Assessment (NR-IQA). In this paper, we hypothesize that latent diffusion models implicitly exhibit perceptually consistent local regions within data manifold. We leverage this insight to guide the on-manifold sampling using perceptual features and input measurements. Specifically, we propose Perceptual Manifold Guidance (PMG), an algorithm that utilizes pretrained latent diffusion models and perceptual quality metrics to obtain perceptually consistent multi-scale and multi-timestep feature maps from the denoising U-Net. We empirically demonstrate that these hyperfeatures exhibit high correlation with human perception in IQA tasks. Our method can be applied to any existing pretrained latent diffusion model and is straightforward to integrate. To the best of our knowledge, this paper is the first work to explore Perceptual Consistency in Diffusion Models (PCDM) and apply it to NR-IQA in a zero-shot setting. Extensive experiments on IQA datasets show that our method, PCDM, achieves state-of-the-art performance, underscoring the superior zero-shot generalization capabilities of diffusion models for NR-IQA tasks.

Primary Area: applications to computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.

Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 9252