Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

An Adaptive Edge-Guided Dual-Network Framework for Fast QR Code Motion Deblurring

Jianping Li, Dongyang Guo, Wenjie Li, Wei Zhao

TL;DR

The paper tackles robust QR code motion deblurring by prioritizing decodability. It introduces Edge-Guided Attention Block (EGAB) to inject explicit edge priors into a Transformer-based Restormer (EG-Restormer) and pairs it with a lightweight LENet in an Adaptive Dual-Network (ADNet) that routes inputs by blur severity using a Laplacian-variance threshold. EG-Restormer delivers state-of-the-art decoding rates while LENet preserves speed for mildly blurred inputs; ADNet balances performance and efficiency with a blur-based routing mechanism and a fallback refinement path. The work demonstrates that domain-specific priors and adaptive architectures can substantially improve QR code restoration on mobile devices, with potential applicability to other structured-text restoration tasks.

Abstract

Unlike general image deblurring that prioritizes perceptual quality, QR code deblurring focuses on ensuring successful decoding. QR codes are characterized by highly structured patterns with sharp edges, a robust prior for restoration. Yet existing deep learning methods rarely exploit these priors explicitly. To address this gap, we propose the Edge-Guided Attention Block (EGAB), which embeds explicit edge priors into a Transformer architecture. Based on EGAB, we develop Edge-Guided Restormer (EG-Restormer), an effective network that significantly boosts the decoding rate of severely blurred QR codes. For mildly blurred inputs, we design the Lightweight and Efficient Network (LENet) for fast deblurring. We further integrate these two networks into an Adaptive Dual-network (ADNet), which dynamically selects the suitable network based on input blur severity, making it ideal for resource-constrained mobile devices. Extensive experiments show that our EG-Restormer and ADNet achieve state-of-the-art performance with a competitive speed. Project page: https://github.com/leejianping/ADNet

An Adaptive Edge-Guided Dual-Network Framework for Fast QR Code Motion Deblurring

TL;DR

The paper tackles robust QR code motion deblurring by prioritizing decodability. It introduces Edge-Guided Attention Block (EGAB) to inject explicit edge priors into a Transformer-based Restormer (EG-Restormer) and pairs it with a lightweight LENet in an Adaptive Dual-Network (ADNet) that routes inputs by blur severity using a Laplacian-variance threshold. EG-Restormer delivers state-of-the-art decoding rates while LENet preserves speed for mildly blurred inputs; ADNet balances performance and efficiency with a blur-based routing mechanism and a fallback refinement path. The work demonstrates that domain-specific priors and adaptive architectures can substantially improve QR code restoration on mobile devices, with potential applicability to other structured-text restoration tasks.

Abstract

Unlike general image deblurring that prioritizes perceptual quality, QR code deblurring focuses on ensuring successful decoding. QR codes are characterized by highly structured patterns with sharp edges, a robust prior for restoration. Yet existing deep learning methods rarely exploit these priors explicitly. To address this gap, we propose the Edge-Guided Attention Block (EGAB), which embeds explicit edge priors into a Transformer architecture. Based on EGAB, we develop Edge-Guided Restormer (EG-Restormer), an effective network that significantly boosts the decoding rate of severely blurred QR codes. For mildly blurred inputs, we design the Lightweight and Efficient Network (LENet) for fast deblurring. We further integrate these two networks into an Adaptive Dual-network (ADNet), which dynamically selects the suitable network based on input blur severity, making it ideal for resource-constrained mobile devices. Extensive experiments show that our EG-Restormer and ADNet achieve state-of-the-art performance with a competitive speed. Project page: https://github.com/leejianping/ADNet
Paper Structure (11 sections, 2 equations, 3 figures, 2 tables, 1 algorithm)

This paper contains 11 sections, 2 equations, 3 figures, 2 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. PROPOSED METHOD
  3. Overall Architecture
  4. Blur Severity-based Routing Unit
  5. Edge-Guided Attention Block
  6. Simple Gate Depthwise Convolution Block
  7. Experiments
  8. Datasets and Metrics
  9. Implementation Details
  10. Results and Discussion
  11. CONCLUSION

Figures (3)

  • Figure 1: Comparison of gradient statistics: natural image vs QR code.
  • Figure 2: The overall architecture of the proposed ADNet framework. (a) The ADNet consists of a powerful EG-Restormer and the LENet. (b) The EGAB uses the EGA module to inject an explicit edge feature $E$ into the attention mechanism. (c) The structure of the EGA module. (d) The architecture of the SGDB. (e) The structure of the edge sharpening attention block (ESAB).
  • Figure 3: Qualitative comparison of QR code deblurring on the test set.