Towards a Handwritten Text Interpretation Framework for Ancient Spanish Manuscripts

Abstract

Handwritten Text Recognition is an extensively studied research topic.We implement a widely known binarization method in order to preprocess handwritten text images efficiently and accurately, acquiring adequate binary black-white images for later recognition processes. Afterwards, the characters present in the documents are used to train and evaluate deep-learning mechanisms for the recognition task. Our framework provides good source images for the recognition phase in terms of noise removal and processing of low contrast images. Besides, the process of character recognition is also improved by means of deep-learning techniques.