Digitalization of Malayalam Palmleaf Manuscripts Based on Contrast-Based Adaptive Binarization and Convolutional Neural Networks

Abstract

The palm leaf manuscripts are an abundant source of knowledge, tradition and ancient culture. These scriptures are an unavoidable part of our rich culture and have to be preserved in the best possible way. But the information extraction from palm leaf is a tedious task due to various challenges such as noise enormous character set and the difficulty in reading and understanding the ancient Malayalam script. Handwriting recognition in Malayalam is a challenging and emerging area of pattern recognition. Our proposed system aims at extracting information from old palm leaves (thaaliyola) and translating the ancient Malayalam scripts to their current version based on contrast-based adaptive binarization and convolutional neural networks which simplifies the entire process by avoiding feature extraction. The proposed method is different from the conventional methods which require handcrafted features that are used for classification. Initially, the system is trained with a set of characters. This can be expanded to work with the remaining characters as well. The input will be images of Malayalam palmleaf manuscript and the expected output is their translated script. Our system aims to transform these scripts so as to make it accessible and useful to the current generation. The system will be trained using a number of samples to build a convolutional neural network using which the characters will be recognized.