SLER-IR: Spherical Layer-wise Expert Routing for All-in-One Image Restoration

2026-03-06

AI Summary

This work addresses the limitations of existing unified image restoration frameworks, which often suffer from feature interference and insufficient expert specialization when handling diverse degradation types. To overcome these issues, the authors propose a spherical hierarchical expert routing mechanism that leverages spherical contrastive learning to construct uniform degradation embeddings, thereby eliminating geometric biases inherent in linear embeddings. Dedicated experts are dynamically activated at each network layer based on the input degradation. Additionally, a global–local granularity fusion module is introduced to effectively handle spatially non-uniform degradations and mitigate discrepancies between training and testing granularities. The proposed method consistently outperforms state-of-the-art approaches on both three-task and five-task restoration benchmarks, achieving notable improvements in PSNR and SSIM metrics.

Technology Category

• Machine Learning: Mixture of Experts (MoE)
• Computer Vision: Adversarial Attacks & Robustness
• Planning, Routing, and Scheduling: Replanning and Plan Repair

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: Graph embeddings and representation learning for Web-related graphs
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Image restoration under diverse degradations remains challenging for unified all-in-one frameworks due to feature interference and insufficient expert specialization. We propose SLER-IR, a spherical layer-wise expert routing framework that dynamically activates specialized experts across network layers. To ensure reliable routing, we introduce a Spherical Uniform Degradation Embedding with contrastive learning, which maps degradation representations onto a hypersphere to eliminate geometry bias in linear embedding spaces. In addition, a Global-Local Granularity Fusion (GLGF) module integrates global semantics and local degradation cues to address spatially non-uniform degradations and the train-test granularity gap. Experiments on three-task and five-task benchmarks demonstrate that SLER-IR achieves consistent improvements over state-of-the-art methods in both PSNR and SSIM. Code and models will be publicly released.

Problem

• image restoration
• all-in-one framework
• feature interference
• expert specialization
• degradation diversity

Innovation

• Spherical Uniform Degradation Embedding
• Expert Routing
• Global-Local Granularity Fusion
• All-in-One Image Restoration
• Contrastive Learning