Watershed transform for text lines extraction on binary Arabic handwriten documents
Abdelghani Souhar, Youssef Boulid, El Bachir Ameur, Mly Moustafa Ouagague · 2017
We propose an approach for automatic text line detection on binary handwritten documents. Our algorithm is based on the Watershed transform to localize components of the same lines. The analysis of the neighbouring regions of each connected component, allow us to link those that satisfy some constraints such as linearity in order to group them in the same line. The preliminary results are promising and show that the proposed approach is language independent.