Preserving the Heritage: Evaluating the Character Segmentation Quality in Palm Leaf Manuscripts by Comparing the Classical and Noise2Void Denoising Techniques
Abstract
The Palm Leaf Manuscripts are a rich source of information about ancient India. It shares an enormous amount of knowledge about the past in terms of art, culture, literature and medicine. As the Manuscripts were developed organically, it is prone to getting damaged very fast. There are many mechanisms used to preserve the physical copies of the manuscripts, but because of the climatic conditions, the deterioration of the manuscripts is inevitable. This work outlines a comparative analysis of classical and deep learning-based approaches for denoising the distorted palm leaf manuscripts based on the segmentation quality of the text inscribed on the PLMs. The traditional pipeline consists of denoising, followed by binarisation and then segmentation of the entire image. We implemented this sequence using both Fast Non-Local Means and a self-trained Noise2Void (N2V) model for denoising. However, the segmented characters, particularly from the Fast NL-based approach, appeared visually distorted. In contrast, the N2V-based difference image showed better structural preservation and closer alignment with the ground truth. To tackle these limitations, we proposed a novel pipeline, which is an innovative processing pipeline that commences with denoising the Palm Leaf Manuscript images using the N2V model, proceeds with direct extraction of the text and culminates in the targeted application of binarisation exclusively on the segmented patches. This restructured approach minimises distortion, enhances text clarity, and preserves character details more effectively. Quantitative evaluation shows improved performance with lower MSE values (0.97, 1.15, 1.02), higher PSNR scores (27.17 dB, 26.61 dB, 29.09 dB) for various binarisation methods, and a structural similarity index (SSIM) of 91%, demonstrating the superiority of the proposed method over the traditional workflow.
Keywords
Binarising, Degraded, Denoising, Distorted, Palm Leaf Manuscripts, SegmentationReferences
[1] Sahoo, J., & Mohanty, B. (2015). Digitization of Indian manuscripts heritage. IFLA Journal, 41(3), 238. DOI: https://doi.org/10.1177/0340035215601447
[2] Suryawanshi, D. G., Nair, M. V., & Sinha, P. M. (1992). Improving the flexibility of palm leaf. Restaurator, 13(1), 37–46. https://doi.org/10.1515/rest.1992.13.1.37 DOI: https://doi.org/10.1515/rest.1992.13.1.37
[3] Davids, B. (2011). From palm leaf to book: A South Asia quest. Printing History, 10(10), 25–37.
[4] Van Dyke, Y. (2009). Sacred leaves: The conservation and exhibition of early Buddhist manuscripts on palm leaves. The Book and Paper Group Annual, 28, 83–97.
[5] Wujastyk, D. (n.d.). Indian manuscripts (p. 159). DOI: https://doi.org/10.1515/9783110225631.159
[6] Wilson, B., & Rice, J. M. (2019). Palm leaf manuscripts in South Asia. School of Information Studies – Post-doc and Student Scholarship, (8). https://surface.syr.edu/ischoolstudents/8
[7] Sulaiman, A., Omar, K., & Nasrudin, M. F. (2019). Degraded historical document binarization: A review on issues, challenges, techniques, and future directions. Journal of Imaging, 5(4), 48. https://doi.org/10.3390/jimaging5040048 DOI: https://doi.org/10.3390/jimaging5040048
[8] Sudarsan, D., & Sankar, D. (2019). A novel approach for denoising palm leaf manuscripts using image gradient approximation. In Proceedings of the IEEE ICECA. DOI: https://doi.org/10.1109/ICECA.2019.8822224
[9] Sudarsan, D., & Sankar, D. (2022). A novel complete denoising solution for old Malayalam palm leaf manuscripts. Pattern Recognition and Image Analysis, 32(1), 187–204. https://doi.org/10.1134/S1054661822010096 DOI: https://doi.org/10.1134/S1054661822010096
[10] Ge, P., Yu, P., Li, H., & Li, H. (2017). Stroke edge based binarization algorithm for the palm leaf manuscripts. In Proceedings of the ICIVC (pp. 778–782). https://doi.org/10.1109/ICIVC.2017.7984660 DOI: https://doi.org/10.1109/ICIVC.2017.7984660
[11] Vats, E., Hast, A., & Singh, P. (2017). Automatic document image binarization using Bayesian optimization. In Proceedings of the International Workshop on Historical Document Imaging and Processing (pp. 89–94). https://doi.org/10.1145/3151509.3151520 DOI: https://doi.org/10.1145/3151509.3151520
[12] Unnikrishnan, D., Sudarsan, D., & Vignesh, R. (2025). A novel method of absolute noise removal from the degraded palm leaf manuscripts. Indian Journal of Traditional Knowledge, 24(6), 595–605. https://doi.org/10.56042/ijtk.v24i6.16782 DOI: https://doi.org/10.56042/ijtk.v24i6.16782
[13] Singh, M., & Indu, S. (2023). Denoising of palm leaf manuscripts using Gaussian filter and conservative smoothing. AIP Conference Proceedings, 2521(1). https://doi.org/10.1063/5.0142237 DOI: https://doi.org/10.1063/5.0142237
[14] Gatos, B., Pratikakis, I., & Perantonis, S. J. (2006). Adaptive degraded document image binarization. Pattern Recognition, 39(3), 317–327. https://doi.org/10.1016/j.patcog.2005.09.010 DOI: https://doi.org/10.1016/j.patcog.2005.09.010
[15] Wagdy, M., Faye, I., & Rohaya, D. (2015). Document Image Binarization Using Retinex and Global Thresholding. ELCVIA Electronic Letters on Computer Vision and Image Analysis, 14(1), 61–73. https://doi.org/10.5565/rev/elcvia.648 DOI: https://doi.org/10.5565/rev/elcvia.648
[16] Jyothi, J., & Malangai, A. (2017). Comparative analysis of wavelet transforms in the recognition of ancient Grantha script. International Journal of Computer Theory and Engineering, 9, 235–241. https://doi.org/10.7763/IJCTE.2017.V9.1144 DOI: https://doi.org/10.7763/IJCTE.2017.V9.1144
[17] Su, B., et al. (2022). A restoration method using dual generative adversarial networks for Chinese ancient characters. Visual Informatics, 6(1), 26–34. DOI: https://doi.org/10.1016/j.visinf.2022.02.001
[18] Uzan, L., Dershowitz, N., & Wolf, L. (2017). Qumran letter restoration by rotation and reflection modified PixelCNN. In Proceedings of the ICDAR, Vol. 1 (pp. 23–29). DOI: https://doi.org/10.1109/ICDAR.2017.14
[19] Chen, K., Seuret, M., Hennebert, J., & Ingold, R. (2017). Convolutional neural networks for page segmentation of historical document images. In Proceedings of the ICDAR, Vol. 1 (pp. 965–970). DOI: https://doi.org/10.1109/ICDAR.2017.161
[20] Pastor-Pellicer, J., Afzal, M. Z., Liwicki, M., & Castro-Bleda, M. J. (2016). Complete system for text line extraction using convolutional neural networks and watershed transform. In Proceedings of the DAS (pp. 30–35). https://doi.org/10.1109/DAS.2016.58 DOI: https://doi.org/10.1109/DAS.2016.58
[21] Renton, G., et al. (2018). Fully convolutional network with dilated convolutions for handwritten text line segmentation. International Journal of Document Analysis and Recognition, 21, 177–186. DOI: https://doi.org/10.1007/s10032-018-0304-3
[22] Alaasam, R., Kurar, B., & El-Sana, J. (2019). Layout analysis on challenging historical Arabic manuscripts using Siamese network. In Proceedings of the ICDAR (pp. 738–742). https://doi.org/10.1109/ICDAR.2019.00123 DOI: https://doi.org/10.1109/ICDAR.2019.00123
[23] Prusty, A., Aitha, S., Trivedi, A., & Sarvadevabhatla, R. K. (2019). Indiscapes: Instance segmentation networks for layout parsing of historical Indic manuscripts. In Proceedings of the ICDAR (pp. 999–1006). DOI: https://doi.org/10.1109/ICDAR.2019.00164
[24] Watanabe, K., et al. (2019). Japanese character segmentation for historical handwritten official documents using fully convolutional networks. In Proceedings of the ICDAR (pp. 934–940). DOI: https://doi.org/10.1109/ICDAR.2019.00154
[25] Ziran, Z., et al. (2020). Text alignment in early printed books combining deep learning and dynamic programming. Pattern Recognition Letters, 133, 109–115. DOI: https://doi.org/10.1016/j.patrec.2020.02.016
[26] Cai, J., Peng, L., Tang, Y., Liu, C., & Li, P. (2019). TH-GAN: Generative adversarial network based transfer learning for historical Chinese character recognition. In Proceedings of the ICDAR (pp. 178–183). DOI: https://doi.org/10.1109/ICDAR.2019.00037
[27] Abbas, A., Baheeja, K., & Alzubaidi, A. M. N. (2023). Ancient textual restoration using deep neural networks: A literature review. In Proceedings of the AICCIT (pp. 64–69). https://doi.org/10.1109/AICCIT57614.2023.10218159 DOI: https://doi.org/10.1109/AICCIT57614.2023.10218159
[28] Bradley, D., & Roth, G. (2007). Adaptive thresholding using the integral image. Journal of Graphics Tools, 12(2), 13–21. https://doi.org/10.1080/2151237x.2007.10129236 DOI: https://doi.org/10.1080/2151237X.2007.10129236
[29] Mittal, A., Soundararajan, R., & Bovik, A. C. (2013). Making a ‘completely blind’ image quality analyzer. IEEE Signal Processing Letters, 20(3), 209–212. https://doi.org/10.1109/LSP.2012.2227726 DOI: https://doi.org/10.1109/LSP.2012.2227726
[30] Venkatanath, N., Praneeth, D., Bh, K. C., Channappayya, S. S., & Medasani, S. S. (2015). Blind image quality evaluation using perception based features. In Proceedings of the NCC (pp. 1–6). https://doi.org/10.1109/NCC.2015.7084921 DOI: https://doi.org/10.1109/NCC.2015.7084843
[31] Mittal, A., Moorthy, A. K., & Bovik, A. C. (2012). No-reference image quality assessment in the spatial domain. IEEE Transactions on Image Processing, 21(12), 4695–4708. https://doi.org/10.1109/TIP.2012.2214050 DOI: https://doi.org/10.1109/TIP.2012.2214050
[32] Sokolova, M., & Lapalme, G. (2009). A systematic analysis of performance measures for classification tasks. Information Processing & Management, 45(4), 427–437. DOI: https://doi.org/10.1016/j.ipm.2009.03.002
[33] Chebbi, E., Benzarti, F., & Amiri, H. (2014). An improvement of structural similarity index for image quality assessment. Journal of Computer Science, 10(2), 353–360. https://doi.org/10.3844/jcssp.2014.353.360 DOI: https://doi.org/10.3844/jcssp.2014.353.360
[34] Gogoi, M., & Ahmed, M. (2016). Image quality parameter detection: A study. International Journal of Computer Science and Engineering, 4, 110–116.
[35] Pang, C., Au, O. C., Dai, J., Yang, W., & Zou, F. (2009). A fast NL-means method in image denoising based on the similarity of spatially sampled pixels. In Proceedings of the IEEE International Workshop on Multimedia Signal Processing, Rio de Janeiro, Brazil (pp. 1–4). DOI: https://doi.org/10.1109/MMSP.2009.5293567
[36] Krull, A., Buchholz, T.-O., & Jug, F. (2019). Noise2Void – Learning denoising from single noisy images. In Proceedings of the CVPR (pp. 2124–2132). https://doi.org/10.1109/CVPR.2019.00223 DOI: https://doi.org/10.1109/CVPR.2019.00223
[37] Otsu, N. (1979). A threshold selection method from gray-level histograms. IEEE Transactions on Systems, Man, and Cybernetics, 9(1), 62–66. https://doi.org/10.1109/tsmc.1979.4310076 DOI: https://doi.org/10.1109/TSMC.1979.4310076
[38] Sauvola, J., & Pietikäinen, M. (2000). Adaptive document image binarization. Pattern Recognition, 33(2), 225–236. https://doi.org/10.1016/S0031-3203(99)00055-2 DOI: https://doi.org/10.1016/S0031-3203(99)00055-2
[39] Chen, D., & Haralick, R. M. (2000). Recognition of degraded characters using weighted shape similarity. Pattern Recognition, 33(7), 1037–1048. https://doi.org/10.1016/S0031-3203(99)00165-6
[40] Reza, A. M. (2004). Realization of the contrast limited adaptive histogram equalization (CLAHE) for real-time image enhancement. Journal of VLSI Signal Processing, 38(1), 35–44. https://doi.org/10.1023/b:vlsi.0000028532.53893.82 DOI: https://doi.org/10.1023/B:VLSI.0000028532.53893.82
Published
How to Cite
Downloads
Copyright (c) 2026 Deepa Unnikrishnan, Vignesh Radhakrishnan
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
