Estimation of Skew Angle from Trilingual Handwritten Documents at Word Level: An Approach Based on Region Props

In this work, an efficient technique for segmenting the words from multilingual unconstrained handwritten documents at word level is proposed. In the proposed model, morphological operations and connected component analysis are used for word identification. Based on that, the bounding box for each word is drawn and then words are segmented. The proposed algorithm also works on documents with any orientation. We conducted experimentation on our own 300 unconstrained multilingual handwritten documents. The result shows the performance of the proposed algorithm.