MiM-DiT: MoE in MoE with Diffusion Transformers for All-in-One Image Restoration

TL;DR

This paper introduces MiM-DiT, a unified image restoration framework combining dual-level MoE architecture with a diffusion model to effectively handle diverse degradation types in a single model.

Contribution

It presents a novel dual-level MoE architecture integrated with a pretrained diffusion model for all-in-one image restoration.

Findings

Outperforms state-of-the-art methods on multiple restoration tasks.

Effectively handles diverse degradation types with high specialization.

Achieves both coarse and fine-grained adaptation within a unified model.

Abstract

All-in-one image restoration is challenging because different degradation types, such as haze, blur, noise, and low-light, impose diverse requirements on restoration strategies, making it difficult for a single model to handle them effectively. In this paper, we propose a unified image restoration framework that integrates a dual-level Mixture-of-Experts (MoE) architecture with a pretrained diffusion model. The framework operates at two levels: the Inter-MoE layer adaptively combines expert groups to handle major degradation types, while the Intra-MoE layer further selects specialized sub-experts to address fine-grained variations within each type. This design enables the model to achieve coarse-grained adaptation across diverse degradation categories while performing fine-grained modulation for specific intra-class variations, ensuring both high specialization in handling complex, real-world corruptions. Extensive experiments demonstrate that the proposed method performs favorably against the state-of-the-art approaches on multiple image restoration task.