# Edge-Efficient Image Restoration: Transformer Distillation into State-Space Models

arXiv: https://arxiv.org/abs/2605.02794 · HTML: https://mlsanity.com/paper/2605.02794/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 44 of its 70 references that are in the map.
- First posted on arXiv 4 May 2026.

## Builds on (most cited first, 30 of 44)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale](https://mlsanity.com/paper/2010.11929/) (2020)
- [Swin Transformer: Hierarchical Vision Transformer using Shifted Windows](https://mlsanity.com/paper/2103.14030/) (2021)
- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://mlsanity.com/paper/2312.00752/) (2023)
- [Longformer: The Long-Document Transformer](https://mlsanity.com/paper/2004.05150/) (2020)
- [SwinIR: Image Restoration Using Swin Transformer](https://mlsanity.com/paper/2108.10257/) (2021)
- [Big Bird: Transformers for Longer Sequences](https://mlsanity.com/paper/2007.14062/) (2020)
- [Restormer: Efficient Transformer for High-Resolution Image Restoration](https://mlsanity.com/paper/2111.09881/) (2021)
- [Rethinking Attention with Performers](https://mlsanity.com/paper/2009.14794/) (2020)
- [Linformer: Self-Attention with Linear Complexity](https://mlsanity.com/paper/2006.04768/) (2020)
- [Pre-Trained Image Processing Transformer](https://mlsanity.com/paper/2012.00364/) (2020)
- [Uformer: A General U-Shaped Transformer for Image Restoration](https://mlsanity.com/paper/2106.03106/) (2021)
- [Multi-Stage Progressive Image Restoration](https://mlsanity.com/paper/2102.02808/) (2021)
- [Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring](https://mlsanity.com/paper/1612.02177/) (2016)
- [Understanding the Effective Receptive Field in Deep Convolutional Neural Networks](https://mlsanity.com/paper/1701.04128/) (2017)
- [CSWin Transformer: A General Vision Transformer Backbone with Cross-Shaped Windows](https://mlsanity.com/paper/2107.00652/) (2021)
- [DeblurGAN: Blind Motion Deblurring Using Conditional Adversarial Networks](https://mlsanity.com/paper/1711.07064/) (2017)
- [DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better](https://mlsanity.com/paper/1908.03826/) (2019)
- [Deep Joint Rain Detection and Removal from a Single Image](https://mlsanity.com/paper/1609.07769/) (2016)
- [Scale-recurrent Network for Deep Image Deblurring](https://mlsanity.com/paper/1802.01770/) (2018)
- [Rethinking Coarse-to-Fine Approach in Single Image Deblurring](https://mlsanity.com/paper/2108.05054/) (2021)
- [Learning Enriched Features for Real Image Restoration and Enhancement](https://mlsanity.com/paper/2003.06792/) (2020)
- [Progressive Image Deraining Networks: A Better and Simpler Baseline](https://mlsanity.com/paper/1901.09221/) (2019)
- [Multi-Scale Progressive Fusion Network for Single Image Deraining](https://mlsanity.com/paper/2003.10985/) (2020)
- [Deep Stacked Hierarchical Multi-patch Network for Image Deblurring](https://mlsanity.com/paper/1904.03468/) (2019)
- [Recurrent Squeeze-and-Excitation Context Aggregation Net for Single Image Deraining](https://mlsanity.com/paper/1807.05698/) (2018)
- [Deblurring by Realistic Blurring](https://mlsanity.com/paper/2004.01860/) (2020)
- [Image Restoration with Mean-Reverting Stochastic Differential Equations](https://mlsanity.com/paper/2301.11699/) (2023)
- [Human-Aware Motion Deblurring](https://mlsanity.com/paper/2001.06816/) (2020)
- [MambaOut: Do We Really Need Mamba for Vision?](https://mlsanity.com/paper/2405.07992/) (2024)

## Abstract (arXiv)

We propose a modular framework for hybrid image restoration that integrates transformer and state-space model (SSM) blocks with a focus on improving runtime efficiency on edge hardware. While transformers provide strong global modeling through self-attention, their attention kernels incur substantial latency on mobile devices, especially for high-resolution inputs. In contrast, SSMs such as Mamba offer lineartime sequence modeling with lower runtime overhead but may underperform on fine grained restoration tasks. To balance accuracy and efficiency, we train lightweight SSM blocks as feature-distilled surrogates of transformer blocks and use them to construct hybrid U-Net-style architectures. To automatically discover effective block combinations, we introduce Efficient Network Search (ENS), a multi-objective search strategy that selects task-specific hybrid configurations from pre-aligned components. ENS optimizes restoration quality while penalizing transformer usage, serving as a lightweight proxy for latency and enabling architecture discovery without repeated hardware profiling. On a Snapdragon 8 Elite CPU, the Restormer baseline requires 10119.52 ms for inference. In contrast, ENS-discovered hybrids significantly reduce runtime: ENS-Deblurring runs in 2973 ms (3.4x faster), ENS-Deraining in 5816 ms (1.74x faster), and ENS-Denoising in 8666 ms (1.17x faster), while maintaining competitive restoration quality.