Text line segmentation using Viterbi algorithm for the palm leaf manuscripts of Dai

Abstract

The text line segmentation process is a key step in an optical character recognition (OCR) system. Several common approaches, such as projection-based methods and stochastic methods, have been put forward to fulfill this task. However, most of existing methods cannot be directly applied to process the palm leaf manuscripts of Dai which the images have poor quality and include smudges, creases, stroke deformation and character touching. To solve this problem, an improved Viterbi algorithm based on Hidden Markov Model (HMM) is proposed to find all possible segmentation paths firstly. And then, a path filtering method is used to detect the optimal paths for the segmented text blocks. The performance of the method is compared with relevant methods and the experimental results demonstrate the effectiveness of the proposed method.