Enhancement of Palm Leaf Manuscripts Images Using Deep SegNet Model

Preserving historical documents is crucial for protecting cultural heritage and ensuring that important information is accessible to future generations. Binarization is crucial in digitizing and restoring degraded manuscripts, particularly palm leaf manuscripts. It helps enhance quality by removing noise, improving contrast, and retaining text. This work proposes a semantic segmentation-based approach for binarizing palm leaf manuscripts using SegNet, a convolutional neural network architecture. Our approach accurately segments text from the background of the manuscript, enabling quality binarization. The suggested method demonstrated an accuracy of about 85% when testing was done on a dataset of palm leaf manuscript images. The method has significant implications for the preservation and digitization of historical documents as palm leaf manuscripts are vulnerable to degradation due to their fragile nature and age. By accurately binarizing palm leaf manuscripts, the approach can help to ensure that these valuable cultural artifacts are accessible to researchers and the public for years to come.