Citation :

Bhagyashri Agasar, Gururaj Mukarambi, "Kannada Stone Inscription Image Enhancement using Modified Approach for Binarization and Detection of Text Regions using Deep Learning YOLO Models," International Journal of Engineering Trends and Technology (IJETT), vol. 74, no. 4, pp. 155-169, 2026. Crossref, https://doi.org/10.14445/22315381/IJETT-V74I4P112

Abstract

In this paper, the YOLOv families, i.e., YOLOv8 to YOLOv11, are used for the detection of TextRegion and Non-textRegion in the stone inscription images. No standard datasets are available in the literature to evaluate the performance of YOLO models on inscription images. Hence, collected 370 inscription images from the ASI Departments, Museums, and ancient temples in the state of Karnataka. The tree structure approach divides each image into four parts based on the contrast of the grayscale image. Then, these four parts undergo contrast enhancement using linear scaling and CLAHE. Further, Otsu's thresholding is applied to greyscale images to convert them to binary images. Then, morphological procedures were applied, including an opening operation to remove minor noise and a closing operation to fill in gaps, resulting in a more enhanced binary image with a PSNR of 35.57 dB. Further, used a set of object detection models based on the YOLO architecture, specifically YOLOv8 to YOLOv11, to distinguish the text and non-text regions. For the experimental setup, the YOLOv (n, s, m, l, x) version has been used. The data augmentation included random horizontal flipping, brightness/contrast jittering, small rotations (±10°), mosaic augmentation, and multi-scale augmentation. For training, used the AdamW optimizer with an initial learning rate in the range 1e-4–1e-3, weight decay in the range 1e-4-1e-3, and batch sizes chosen according to GPU memory constraints; final hyperparameters were selected by empirical validation sweeps (learning rate = 1e-4, weight decay = 5e-4, batch = 32). To evaluate the performance of the YOLO models, precision, recall, F1-score, and accuracy are used. In the experimental results, YOLOv11 performs better than other versions, such as YOLOv8, v9, and v10. The YOLOv11s achieved the highest accuracy of 85.20%. This will help advance the larger goal of preserving and interpreting digitized images of stone inscriptions.

Keywords

Inscriptions, YOLO, CLAHE, Binary images, Deep learning.

References

[1] Shashaank M. Aswatha et al., “A Method for Extracting Text from Stone Inscriptions using Character Spotting,” Computer Vision - ACCV 2014 Workshops: Singapore, Singapore, vol. 9009, pp. 598-611, 2015.
[CrossRef] [Google Scholar] [Publisher Link]

[2] Sachin Bhat, and G Seshikala, “Preprocessing and Binarisation of Inscription Images using Phase-based Features,” 2018 Second International Conference on Advances in Electronics, Computers and Communications (ICAECC), Bangalore, India, pp. 1-6, 2018.
[CrossRef] [Google Scholar] [Publisher Link]

[3] S. Bhuvaneswari, and K. Kathiravan, “Enhancing Epigraphy: A Deep Learning Approach to Recognize and Analyze Tamil Ancient Inscriptions,” Neural Computing and Applications, vol. 36, no. 31, pp. 19839-19861, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[4] N. Shobha Rani, and Arun Gopi, “A Quad Tree based Binarization Approach to Improve Quality of Degraded Document Images,” International Journal of Computer Science Engineering (IJCSE), vol. 3, no. 1, pp. 1-8, 2024.
[Google Scholar] [Publisher Link]

[5] Miguel Carrero-Pazos, and David Espinosa-Espinosa, “Tailoring 3D Modelling Techniques for Epigraphic Texts Restitution: Case Studies in Deteriorated Roman Inscriptions,” Digital Applications in Archaeology and Cultural Heritage, vol. 10, pp. 1-28, 2018.
[CrossRef] [Google Scholar] [Publisher Link]

[6]  H.T. Chandrakala, G. Thippeswamy, and Roshan Joy Martis, “Impact of Total Variation Regularization on Character Segmentation from Historical Stone Inscriptions,” Pattern Recognition and Image Analysis, vol. 31, no. 1, pp. 35-48, 2021.
[CrossRef] [Google Scholar] [Publisher Link]

[7] Ushasi Chaudhuri, Partha Bhowmick, and Jayanta Mukherjee, “A Novel Rough Set based Technique for Character Spotting on Inscription Images,” 2017 Ninth International Conference on Advances in Pattern Recognition (ICAPR), Bangalore, India, pp. 1-6, 2017.
[CrossRef] [Google Scholar] [Publisher Link]

[8] Sugata Das, Sekhar Mandal, and Amit Kumar Das, “Binarization of Stone Inscripted Documents,” 2015 IEEE International Conference on Computer Graphics, Vision and Information Security (CGVIS), Bhubaneswar, India, pp. 11-16, 2015.
[CrossRef] [Google Scholar] [Publisher Link]

[9] K. Durga Devi et al., “Pattern Matching Model for Recognition of Stone Inscription Characters,” The Computer Journal, vol. 66, no. 3, pp. 554-564, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[10] K. Durga Devi, and P. Uma Maheswari, “RETRACTED ARTICLE: Digital Acquisition and Character Extraction from Stone Inscription Images using Modified Fuzzy Entropy-based Adaptive Thresholding,” Soft Computing, vol. 23, no. 8, pp. 2611-2626, 2018.
[CrossRef] [Google Scholar] [Publisher Link]

[11] Rafael C. Gonzalez, and Richared E. Woods, Digital Image Processing, 3^rd ed., Pearson, 2009.
[Google Scholar]

[12] J. Jayanthi, and P. Uma Maheswari, “Comparative Study: Enhancing Legibility of Ancient Indian Script Images from Diverse Stone Background Structures using 34 Different Pre-Processing Methods,” Heritage Science, vol. 12, no. 1, pp. 1-17, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[13] Ke Liu, and Jun Ma, “3D-Scanning and Computer Reverse Engineering Technology to Preserve Inscriptions at Beihai Park,” International Journal of Simulation: Systems, Science and Technology, vol. 17, no. 26, 2016.
[CrossRef] [Google Scholar] [Publisher Link]

[14] H.S. Mohana et al., “Interactive Segmentation for Character Extraction in Stone Inscriptions,” Second International Conference on Current Trends in Engineering and Technology - ICCTET 2014, Coimbatore, pp. 321-327, 2014.
[CrossRef] [Google Scholar] [Publisher Link]

[15] Monisha Munivel, and V.S. Felix Enigo, “MLIBT: A Multi-Level Improvised Binarization Technique for Tamizhi Inscriptions,” Expert Systems with Applications, vol. 236, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[16] Pranav Rajnish et al., “Improving the Quality and Readability of Ancient Brahmi Stone Inscriptions,” 2023 2^nd International Conference for Innovation in Technology (INOCON), Bangalore, India, pp. 1-8, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[17] N. Sasipriyaa et al., “An Approach for Tamil Handwritten Recognition using ABC-Enabled GAN,” 2024 2^nd International Conference on Signal Processing, Communication, Power and Embedded System (SCOPES), Paralakhemundi Campus, Centurion University of Technology and Management, Odisha, India, pp. 1-5, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[18] Indu Sreedevi et al., “NGFICA based Digitization of Historic Inscription Images,” International Scholarly Research Notices, vol. 2013, no. 1, pp. 1-7, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[19] S. Sukanthi, S. Sakthivel Murugan, and S. Hanis, “Binarization of Stone Inscription Images by Modified Bi-Level Entropy Thresholding,” Fluctuation and Noise Letters, vol. 20, no. 6, 2021.
[CrossRef] [Google Scholar] [Publisher Link]

[20] Ultralytics, Explore Ultralytics YOLOv8, 2023. [Online]. Available: https://docs.ultralytics.com/models/yolov8/

[21] Ultralytics, YOLOv9 vs. YOLO11: A Technical Deep Dive into Modern Object Detection, 2024. [Online]. Available: https://docs.ultralytics.com/compare/yolov9-vs-yolo11/

[22] Karel Zuiderveld, “Contrast Limited Adaptive Histogram Equalization,” Graphics Gems IV, pp. 474-485, 1994.
[Google Scholar] [Publisher Link]

[23] Bapu D. Chendage, and Rajivkumar S. Mente, “Enhancement of Ancient Marathi Script using Improved Binarization Method,” Sādhanā, vol. 48, no. 4, pp. 1-5, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[24] Bapu D. Chendagel, Rajivkumar S. Mente, and Bapu D. Chendage, “A Comparative Study of Contrast Enhancement and Brightness Preservation of Ancient Inscription Image using Different Histogram Equalization Algorithms,” International Journal of Natural Sciences, vol. 14, no. 80, pp. 61336-61343, 2023.
[Google Scholar]

[25] A.I. Papadaki et al., “Accurate 3D Scanning of Damaged Ancient Greek Inscriptions for Revealing Weathered Letters,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 40, pp. 237-244, 2015.
[CrossRef] [Google Scholar] [Publisher Link]

[26] Zhongming Pei, Yong Mao Huang, and Ting Zhou, “Review on Analysis Methods Enabled by Hyperspectral Imaging for Cultural Relic Conservation,” Photonics, vol. 10, no. 10, pp. 1-19, 2023.
[CrossRef] [Google Scholar] [Publisher Link]

[27] Haiqing Yang et al., “Hyperspectral Data Set of Stone Cultural Relics in High-Precision Machine Vision Scene,” Scientific Data, vol. 12, no. 1, pp. 1-10, 2025.
[CrossRef] [Google Scholar] [Publisher Link]

[28] Balasubramanian Murugan, and P. Visalakshi, “Ancient Tamil Inscription Recognition using Detect, Recognize and Labelling, Interpreter Framework of Text Method,” Heritage Science, vol. 12, no. 1, pp. 1-21, 2024.
[CrossRef] [Google Scholar] [Publisher Link]

[30] Boris Sekachev, Andrey Zhavoronkov, and Nikita Manovich, “Computer Vision Annotation Tool: A Universal Approach to Data Annotation,” Intel [Internet], 2019.
[Google Scholar]