State of the Field
Recent advancements in image restoration are focusing on enhancing model efficiency and adaptability for real-world applications. Techniques such as Quantization-aware Distilled Restoration are addressing the challenges of deploying high-performance models on edge devices by optimizing for both reconstruction quality and computational efficiency. Meanwhile, frameworks like QualiTeacher are transforming the use of pseudo-labels in training, allowing models to learn from imperfect data without inheriting artifacts, thus improving generalization. The introduction of spherical layer-wise expert routing in SLER-IR is enabling dynamic specialization within networks, which enhances performance across diverse degradation types. Additionally, universal restoration approaches are evolving to support scalable and controllable restoration across multiple degradation scenarios, addressing the limitations of existing all-in-one models. These developments suggest a concerted effort in the field to create robust, efficient, and versatile image restoration solutions that can be effectively deployed in various commercial contexts, from consumer electronics to autonomous systems.
Papers
1–7 of 7Decoder-Free Distillation for Quantized Image Restoration
Quantization-Aware Training (QAT), combined with Knowledge Distillation (KD), holds immense promise for compressing models for edge deployment. However, joint optimization for precision-sensitive imag...
UnSCAR: Universal, Scalable, Controllable, and Adaptable Image Restoration
Universal image restoration aims to recover clean images from arbitrary real-world degradations using a single inference model. Despite significant progress, existing all-in-one restoration networks d...
Derain-Agent: A Plug-and-Play Agent Framework for Rainy Image Restoration
While deep learning has advanced single-image deraining, existing models suffer from a fundamental limitation: they employ a static inference paradigm that fails to adapt to the complex, coupled degra...
UNet-AF: An alias-free UNet for image restoration
The simplicity and effectiveness of the UNet architecture makes it ubiquitous in image restoration, image segmentation, and diffusion models. They are often assumed to be equivariant to translations, ...
QualiTeacher: Quality-Conditioned Pseudo-Labeling for Real-World Image Restoration
Real-world image restoration (RWIR) is a highly challenging task due to the absence of clean ground-truth images. Many recent methods resort to pseudo-label (PL) supervision, often within a Mean-Teach...
SLER-IR: Spherical Layer-wise Expert Routing for All-in-One Image Restoration
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 l...
Progressive Split Mamba: Effective State Space Modelling for Image Restoration
Image restoration requires simultaneously preserving fine-grained local structures and maintaining long-range spatial coherence. While convolutional networks struggle with limited receptive fields, an...