Xuexin Liang , Lu Lu , Qingyan Tian , Haishan Lin , Quanyi Zou
{"title":"TunnelDiff: A brightness-guided image restoration diffusion model for enhancing defect detection in low-quality tunnel lining images","authors":"Xuexin Liang , Lu Lu , Qingyan Tian , Haishan Lin , Quanyi Zou","doi":"10.1016/j.dsp.2025.105581","DOIUrl":null,"url":null,"abstract":"<div><div>During capturing in the tunnel, factors such as lining materials, illumination conditions, and imaging equipment may affect the quality of images and introduce noise. The low-quality images bring challenges in tunnel lining defect detection. This paper introduces TunnelDiff, a diffusion model designed to enhance tunnel images and perform better on defect detection. TunnelDiff restores image details by leveraging the inherent generalization ability of the pretrained Stable Diffusion Model. It also introduces the Condition Module to guide the generation direction. The Condition Module includes the Illumination Distribution Module (IDM) and the Brightness Guided Module (BGM). The IDM focuses on correcting uneven illumination in tunnel images, while the BGM addresses target brightness ambiguity in low-light correction tasks. Due to the absence of paired data in tunnel enhancement, TunnelDiff first trained on the Exposure Errors dataset and then enhanced the Tunnel Defect dataset. Experimental outcomes demonstrated improvements in image quality metrics on both datasets, and the Tunnel Defect dataset enhanced by TunnelDiff performed better in defect detection than datasets enhanced by other models. In particular, TunnelDiff showed better crack defect detection, with 2.03 %, 1.42 %, and 1.55 % improvement in crack recall, F1-score, and IoU. Additionally, TunnelDiff consistently produced images within a specific brightness range. These results underscore the effectiveness of TunnelDiff. The corresponding code is available at: <span><span>https://github.com/derolol/tunnel_diff.git</span><svg><path></path></svg></span>.</div></div>","PeriodicalId":51011,"journal":{"name":"Digital Signal Processing","volume":"168 ","pages":"Article 105581"},"PeriodicalIF":3.0000,"publicationDate":"2025-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Digital Signal Processing","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1051200425006037","RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, ELECTRICAL & ELECTRONIC","Score":null,"Total":0}
引用次数: 0
Abstract
During capturing in the tunnel, factors such as lining materials, illumination conditions, and imaging equipment may affect the quality of images and introduce noise. The low-quality images bring challenges in tunnel lining defect detection. This paper introduces TunnelDiff, a diffusion model designed to enhance tunnel images and perform better on defect detection. TunnelDiff restores image details by leveraging the inherent generalization ability of the pretrained Stable Diffusion Model. It also introduces the Condition Module to guide the generation direction. The Condition Module includes the Illumination Distribution Module (IDM) and the Brightness Guided Module (BGM). The IDM focuses on correcting uneven illumination in tunnel images, while the BGM addresses target brightness ambiguity in low-light correction tasks. Due to the absence of paired data in tunnel enhancement, TunnelDiff first trained on the Exposure Errors dataset and then enhanced the Tunnel Defect dataset. Experimental outcomes demonstrated improvements in image quality metrics on both datasets, and the Tunnel Defect dataset enhanced by TunnelDiff performed better in defect detection than datasets enhanced by other models. In particular, TunnelDiff showed better crack defect detection, with 2.03 %, 1.42 %, and 1.55 % improvement in crack recall, F1-score, and IoU. Additionally, TunnelDiff consistently produced images within a specific brightness range. These results underscore the effectiveness of TunnelDiff. The corresponding code is available at: https://github.com/derolol/tunnel_diff.git.
期刊介绍:
Digital Signal Processing: A Review Journal is one of the oldest and most established journals in the field of signal processing yet it aims to be the most innovative. The Journal invites top quality research articles at the frontiers of research in all aspects of signal processing. Our objective is to provide a platform for the publication of ground-breaking research in signal processing with both academic and industrial appeal.
The journal has a special emphasis on statistical signal processing methodology such as Bayesian signal processing, and encourages articles on emerging applications of signal processing such as:
• big data• machine learning• internet of things• information security• systems biology and computational biology,• financial time series analysis,• autonomous vehicles,• quantum computing,• neuromorphic engineering,• human-computer interaction and intelligent user interfaces,• environmental signal processing,• geophysical signal processing including seismic signal processing,• chemioinformatics and bioinformatics,• audio, visual and performance arts,• disaster management and prevention,• renewable energy,