ICM-SR: Image-Conditioned Manifold Regularization for Image Super-Resoultion

2025-11-26

AI Summary

Existing real-world image super-resolution (Real-ISR) methods heavily rely on text-conditioned diffusion models, whose generative manifolds misalign with authentic image degradation processes—leading to color distortion and edge blurring. To address this, we propose Image-Conditioned Manifold Regularization (ICMR), the first approach to explicitly embed structural priors—such as chromatic maps and Canny edges—into the diffusion process, thereby constructing a task-aligned, sparse conditional generation manifold. ICMR avoids distortions and instability caused by direct text or low-quality (LQ) image guidance, requires no additional training, and is compatible with mainstream diffusion architectures. Experiments demonstrate that ICMR significantly improves perceptual quality and reconstruction stability across multiple Real-ISR benchmarks, with notable gains in texture fidelity and color consistency. This work establishes an efficient and robust regularization paradigm for real-scenario image super-resolution.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Learning with Manifolds
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Real world image super-resolution (Real-ISR) often leverages the powerful generative priors of text-to-image diffusion models by regularizing the output to lie on their learned manifold. However, existing methods often overlook the importance of the regularizing manifold, typically defaulting to a text-conditioned manifold. This approach suffers from two key limitations. Conceptually, it is misaligned with the Real-ISR task, which is to generate high quality (HQ) images directly tied to the low quality (LQ) images. Practically, the teacher model often reconstructs images with color distortions and blurred edges, indicating a flawed generative prior for this task. To correct these flaws and ensure conceptual alignment, a more suitable manifold must incorporate information from the images. While the most straightforward approach is to condition directly on the raw input images, their high information densities make the regularization process numerically unstable. To resolve this, we propose image-conditioned manifold regularization (ICM), a method that regularizes the output towards a manifold conditioned on the sparse yet essential structural information: a combination of colormap and Canny edges. ICM provides a task-aligned and stable regularization signal, thereby avoiding the instability of dense-conditioning and enhancing the final super-resolution quality. Our experiments confirm that the proposed regularization significantly enhances super-resolution performance, particularly in perceptual quality, demonstrating its effectiveness for real-world applications. We will release the source code of our work for reproducibility.

Problem

• Corrects misalignment in Real-ISR by using image-conditioned manifold regularization
• Addresses instability from dense image conditioning with sparse structural information
• Enhances super-resolution quality by avoiding color distortions and blurred edges

Innovation

• Image-conditioned manifold regularization for super-resolution
• Sparse structural information from colormap and edges
• Enhances perceptual quality and stability in output