Efficient Image Restoration via Latent Consistency Flow Matching

2025-02-05

AI Summary

To address the deployment challenges of generative image restoration models on edge devices, this paper proposes a lightweight and efficient paradigm: first estimating the minimum mean squared error (MMSE) latent representation in the latent space, then mapping it to high-fidelity images via Latent Consistency Flow Matching. This work pioneers the deep integration of latent-space MMSE estimation with consistency-based flow modeling. The resulting framework achieves an optimal trade-off between distortion and perceptual quality while accelerating inference by over 4× and reducing model parameters by 4×, significantly lowering memory and computational overhead. Extensive experiments demonstrate state-of-the-art performance across diverse image restoration tasks—including denoising, deblurring, and super-resolution—validating its effectiveness and efficiency. The method provides a practical, deployable solution for generative restoration under stringent resource constraints.

Technology Category

• Machine Learning: Learning on the Edge & Model Compression
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Search and Retrieval-Augmented AI: Large language models for search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Recent advances in generative image restoration (IR) have demonstrated impressive results. However, these methods are hindered by their substantial size and computational demands, rendering them unsuitable for deployment on edge devices. This work introduces ELIR, an Efficient Latent Image Restoration method. ELIR operates in latent space by first predicting the latent representation of the minimum mean square error (MMSE) estimator and then transporting this estimate to high-quality images using a latent consistency flow-based model. Consequently, ELIR is more than 4x faster compared to the state-of-the-art diffusion and flow-based approaches. Moreover, ELIR is also more than 4x smaller, making it well-suited for deployment on resource-constrained edge devices. Comprehensive evaluations of various image restoration tasks show that ELIR achieves competitive results, effectively balancing distortion and perceptual quality metrics while offering improved efficiency in terms of memory and computation.

Problem

• Efficient image restoration
• Reduced computational demands
• Suitable for edge devices

Innovation

• Latent space image restoration
• Latent consistency flow model
• Efficient edge device deployment