Belge Kurtarmada Derin Öğrenme Yaklaşımı: DenoiseU-Net ile Yüksek Performans

Year 2025, Volume: 25 Issue: 6, 1323 - 1335

Salih Can Turan , Zeki Çıplak , Ali Sarıkaş , Kazım Yıldız

Abstract

Görüntü işlemede önemli bir görev olan görüntü gürültü azaltma, devam eden araştırmalara rağmen sürekli olarak zorluklarla karşılaşmaktadır. Bu araştırmada veri kümesi, bazıları taranmış veya dijitalleştirilmiş belgelerden oluşan 60 kamu kaynağından 20.000 görüntünün çıkarılmasıyla oluşturulmuştur. Her görüntü en az bir düz metin, resim, tablo veya matematiksel ifade içerip içermediğine bakılarak kontrol edildi. Görüntülere rastgele siyah ve beyaz pikseller, Gauss bulanıklığı, gri alanlar, benek gürültüsü, rastgele yönlü çizgiler, Poisson gürültüsü ve tuz ve biber gürültüsü gibi yaygın gürültüler uygulanmıştır. Test setini oluşturmak için, yedi gürültü türü, 3500 görüntüden oluşan dengeli bir test seti oluşturmak üzere veri setinden çıkarılan 500 görüntüye ayrı ayrı eklenmiştir. Veri kümesi, eğitim ve test kümesi oranı 5:1 olan 23.000 görüntüden oluşmaktadır. DenoiseU-Net modelimiz özellikle gürültülü taranmış belgelerin kurtarılmasını hedeflemekte ve tablolar, resimler, matematiksel denklemler ve metin gibi çeşitli içerik türlerinde başarılı olmaktadır. Deneysel sonuçlar, DenoiseU-Net'in test setindeki ortalama kesinlik, geri çağırma ve F1-skorunun sırasıyla %99.36, %99.59 ve %99.48 olduğunu göstermektedir. Bu değerlendirme sonuçlarına ek olarak, görüntülerin kalitesini belirtmek için yaygın olarak kullanılan parametreler olan ortalama SSIM ve PSNR değerleri sırasıyla 0.9657 ve 40.28 dB olarak elde edilmiştir. Genel sonuçlara göre, önerilen DenoiseU-Net yöntemi üstün performans sergilemektedir.

Keywords

Belge kurtarma , taranmış belge , belge görüntüsü gürültü giderme , gürültülü görüntüler , derin öğrenme , yapısal benzerlik indeksi

A Deep Learning Approach to Document Recovery: High Performance with DenoiseU-Net

Abstract

Image denoising, a crucial task in image processing, has consistently faced challenges despite ongoing research efforts. In this research, the dataset was created by extracting 20,000 images from 60 public sources, some of scanned or digitized documents. Each image was checked to see if it contained at least one plain text, image, table or mathematical expression. Common noises such as random black and white pixels, Gaussian blur, gray areas, speckle noise, random directional lines, Poisson noise and salt and pepper noise were applied to the images. To create the test set, the seven types of noise were individually added to 500 images excluded from the dataset to create a balanced test set of 3500 images. The dataset consists of 23,000 images with a training and test set ratio of 5:1. In particular, our DenoiseU-Net model targets the recovery of noisy scanned documents and achieves capability on various content types such as tables, images, mathematical equations, and text. Experimental results show that the average precision, recall and F1-score of DenoiseU-Net on the test set are 99.36%, 99.59% and 99.48%, respectively. In addition to these evaluation results, the average SSIM and PSNR values, which are commonly used parameters to indicate the quality of the images, were obtained as 0.9657 and 40.28 dB, respectively. From overall results, proposed DenoiseU-Net method shows the superior performance.

Keywords

Document recovery , scanned document , document image denoising , noisy images , deep learning , structural similarity index

References

• Alshathri, S. I., Vincent, D. J., & Hari, V. S. (2022). Denoising Letter Images from Scanned Invoices Using Stacked Autoencoders. Computers, Materials & Continua, 71(1), 1371-1386. https://doi.org/10.32604/cmc.2022.022458
• Berkner, K. (2001). Enhancement of scanned documents in Besov spaces using wavelet domain representations. Document Recognition and Retrieval IX. San Jose, CA, USA, 143–154.
• Buades, A., Coll, B., & Morel, J. M. (2005). A non-local algorithm for image denoising. IEEE computer society conference on computer vision and pattern recognition San Diego, CA, USA, 60-65.
• Dong, C., Loy, C. C., He, K., & Tang, X. (2015). Image super-resolution using deep convolutional networks. IEEE transactions on pattern analysis and machine intelligence, 38(2), 295-307. https://doi.org/10.1109/TPAMI.2015.2439281
• Dong, C., Loy, C. C., He, K., & Tang, X. (2014). Learning a deep convolutional network for image super-resolution. 13th European Conference on Computer Vision (ECCV 2014). Zurich, Switzerland, 184-199. Gupta, S. K., Pal, R., Ahmad, A., Melandsø, F., & Habib, A. (2023). Image denoising in acoustic microscopy using block-matching and 4D filter. Scientific Reports, 13, 13212. https://doi.org/10.1038/s41598-023-40301-7
• He, W., Zhang, H., Shen, H., & Zhang, L. (2018). Hyperspectral image denoising using local low-rank matrix recovery and global spatial–spectral total variation. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 11(3), 713-729. https://doi.org/10.1109/JSTARS.2018.2795093
• Hu, L., Hu, Z., Bauer, P., Harris, T. J., & Allebach, J. P. (2021). Deep learning approaches to determining optimal resolution for scanned text documents. Electronic Imaging, 33, 1-8. https://doi.org/10.2352/ISSN.2470-1173.2021.16.COLOR-243
• Huang, J. B., Singh, A., & Ahuja, N. (2015). Single image super-resolution from transformed self-exemplars. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Boston, MA, USA, 5197-5206.
• Hsu, E., Malagaris, I., Kuo, Y. F., Sultana, R., & Roberts, K. (2022). Deep learning-based NLP data pipeline for EHR-scanned document information extraction. JAMIA open, 5(2), 1-12. https://doi.org/10.1093/jamiaopen/ooac045
• Jadhav, P., Sawal, M., Zagade, A., Kamble, P., & Deshpande, P. (2022) . Pix2pix generative adversarial network with resnet for document image denoising. 4th International Conference on Inventive Research in Computing Applications (ICIRCA). Coimbatore, India, 1489-1494.
• Jiang, Y., Xia, Y., Feng, Z., & Li, J. (2017, July). Research on the recovery technology of scanned image of obsolete document. 4th International Conference on Information, Cybernetics and Computational Social Systems (ICCSS), Dalian, China, 2017, 61-65.
• Kim, J., Lee, J. K., & Lee, K. M. (2016). Accurate image super-resolution using very deep convolutional networks. Proceedings of the IEEE conference on computer vision and pattern recognition, Las Vegas Nevada, 1646-1654.
• Kreuzer, D., & Munz, M. (2023). Transformer-Based UNet with Multi-Headed Cross-Attention Skip Connections to Eliminate Artifacts in Scanned Documents. arXiv preprint arXiv:2306.02815. https://doi.org/10.48550/arXiv.2306.02815
• Kulkarni, M., Kakad, S., Mehra, R., & Mehta, B. (2020). Denoising documents using image processing for digital restoration. Proceedings of ICMLIP Machine Learning and Information Processing. Singapore, 287-295.
• Liu, W., Li, Y., & Huang, D. (2023). RA-UNet: an improved network model for image denoising. The Visual Computer, 40(6), 4319-4335. https://doi.org/10.1007/s00371-023-03084-6
• Mange, G., Mwangi, W., Kimwele, M., & Gómez, J. M. (2023). A Cnn Model for Improved Image Denoising with an Attention Guided Feature Selection. 20 August 2023, PREPRINT (Version 1) https://doi.org/10.21203/rs.3.rs-3267082/v1
• Mehta, D., Padalia, D., Vora, K., & Mehendale, N. (2022, December). MRI image denoising using U-Net and Image Processing Techniques. 5th International Conference on Advances in Science and Technology (ICAST), Mumbai, India, 306-313.
• Mikołajczyk, A., & Grochowski, M. (2018, May). Data augmentation for improving deep learning in image classification problem. International interdisciplinary PhD workshop, Poland, 117-122.
• Moghadam, F. S., & Rashidi, S. (2024). Novel feature extraction based on DCT-DOST features for classification of Digital Breast Tomosynthesis images into benign and malignant tumors. 07 February 2024, PREPRINT (Version 1), available at Research Square https://doi.org/10.21203/rs.3.rs-3931625/v1
• Najafabadi, M. M., Villanustre, F., Khoshgoftaar, T. M., Seliya, N., Wald, R., & Muharemagic, E. (2015). Deep learning applications and challenges in big data analytics. Journal of big data, 2, 1-21 https://doi.org/10.1186/s40537-014-0007-7
• Oktay, O., Schlemper, J., Folgoc, L. L., Lee, M., Heinrich, M., Misawa, et al. (2018). Attention u-net: Learning where to look for the pancreas. arXiv preprint arXiv:1804.03999. https://doi.org/10.48550/arXiv.1804.03999
• Paliwal, S. S., Vishwanath, D., Rahul, R., Sharma, M., & Vig, L. (2019, September). Tablenet: Deep learning model for end-to-end table detection and tabular data extraction from scanned document images. International Conference on Document Analysis and Recognition (ICDAR), Sydney, NSW, Australia, 128-133.
• Patel, K. K., Goyal, S. K., & Patel, Y. K. (2023). Image Processing for Food Safety and Quality. Novel Technologies in Food Science, Navnidhi Chhikara, Anil Panghal, Gaurav Chaudhary (Editors), Wiley, 451-478.
• Rafiee, A. A., & Farhang, M. (2023). A deep convolutional neural network for salt-and-pepper noise removal using selective convolutional blocks. Applied Soft Computing, 145, 110535. https://doi.org/10.1016/j.asoc.2023.110535
• Rashmi, R., Prasad, K., & Udupa, C. B. K. (2022). Breast histopathological image analysis using image processing techniques for diagnostic purposes: A methodological review. Journal of Medical Systems, 46(7), https://doi.org/10.1007/s10916-021-01786-9
• Ronneberger, O., Fischer, P., & Brox, T. (2015). U-net: Convolutional networks for biomedical image segmentation. 18th international conference on medical image computing and computer-assisted intervention, Munich, Germany, 234-241.
• Rudin, L. I., Osher, S., & Fatemi, E. (1992). Nonlinear total variation based noise removal algorithms. Physica D: nonlinear phenomena, 60(1-4), 259-268. https://doi.org/10.1016/0167-2789(92)90242-F
• Salvi, M., Acharya, U. R., Molinari, F., & Meiburger, K. M. (2021). The impact of pre-and post-image processing techniques on deep learning frameworks: A comprehensive review for digital pathology image analysis. Computers in Biology and Medicine, 128, 104129. https://doi.org/10.1016/j.compbiomed.2020.104129
• Sezer, A., & Altan, A. (2021). Detection of solder paste defects with an optimization‐based deep learning model using image processing techniques. Soldering & Surface Mount Technology, 33(5), 291-298. https://doi.org/10.1108/SSMT-04-2021-0013
• Srinivasa, K. G., Sowmya, B. J., Kumar, D. P., & Shetty, C. (2016). Efficient Image Denoising for Effective Digitization using Image Processing Techniques and Neural Networks. International Journal of Applied Evolutionary Computation, 7(4), 77-93. https://doi.org/10.4018/IJAEC.2016100105
• Schreiber, S., Agne, S., Wolf, I., Dengel, A., & Ahmed, S. (2017). Deepdesrt: Deep learning for detection and structure recognition of tables in document images. 14th IAPR international conference on document analysis and recognition (ICDAR), Kyoto, Japan, 1162-1167.
• Tahir, H., & Din, A. H. M. (2024). The Potential of Landsat 8 OLI Images in Coastline Identification: The Case Study of Basra, Iraq. Engineering, Technology & Applied Science Research, 14(1), 13041-13046. https://doi.org/10.48084/etasr.6580
• Tomasi, C., & Manduchi, R. (1998). Bilateral filtering for gray and color images. Sixth international conference on computer vision, Bombay, India, 839-846.
• Vincent, P., Larochelle, H., Bengio, Y., & Manzagol, P. A. (2008). Extracting and composing robust features with denoising autoencoders. In Proceedings of the 25th international conference on Machine learning, Helsinki, Finland, 1096-1103.
• Wang, Z., Bovik, A. C., Sheikh, H. R., & Simoncelli, E. P. (2004). Image quality assessment: from error visibility to structural similarity. IEEE transactions on image processing, 13(4), 600-612. https://doi.org/10.1109/TIP.2003.819861
• Yin, J., Xu, K., Kan, J., Dong, F., & Chen, K. (2023). A Dual-U Structure for Image Denoising Based on Attention Mechanism. 2nd International Conference on Computing, Communication, Perception and Quantum Technology (CCPQT), Xiamen, China, 421-426.
• Zhao, R., Lun, D. P., & Lam, K. M. (2020). Enhancing and Learning Denoiser without Clean Reference. arXiv preprint arXiv:2009.04286. https://doi.org/10.48550/arXiv.2009.04286
• Zhang, K., Zuo, W., Chen, Y., Meng, D., & Zhang, L. (2017). Beyond a gaussian denoiser: Residual learning of deep cnn for image denoising. IEEE transactions on image processing, 26(7), 3142-3155. https://doi.org/10.1109/TIP.2017.2662206
• Zhang, J., Niu, Y., Shangguan, Z., Gong, W., & Cheng, Y. (2023). A novel denoising method for CT images based on U-net and multi-attention. Computers in Biology and Medicine, 152, 106387. https://doi.org/10.1016/j.compbiomed.2022.106387
• Zhang, Q., Xiao, J., Tian, C., Chun‐Wei Lin, J., & Zhang, S. (2023). A robust deformed convolutional neural network (CNN) for image denoising. CAAI Transactions on Intelligence Technology, 8(2), 331-342. https://doi.org/10.1049/cit2.12110
• Zulkarnain, I., Nurmalasari, R. R., & Azizah, F. N. (2022). Table information extraction using data augmentation on deep learning and image processing. 16th International Conference on Telecommunication Systems, Services, and Applications, Lombok, Indonesia, 1-6.