Removal of degradation from ancient stone inscription images using deep learning techniques
Bipin Nair B J, P M Veiyo, Deepti Gupta, Vinayakumar Ravi, G S Vijay, Avula Sachin, S. Akhil
Abstract
Important knowledge, wisdom, and language are lost over time due to fading, erosion, and environmental variables that complicate ancient stone inscriptions. To help with protection efforts, the proposed study employs an image binarization model based on well-known deep neural networks algorithms. The methods durability is enhanced, and a more effective stone engraving preservation strategy is ensured by the application of multiple deep neural networks architectures. This research presents a model that has attained an incredible average 90.76% accuracy. This was made possible by extensive training with 437 images from temples in Karnataka. The deep neural networks technique is used to enhance the efficiency and accuracy of stone inscription preservation while reducing the challenges caused by environmental factors such as algae.