Recognition of Tibetan wood block prints with generalized hidden Markov and kernelized modified quadratic distance function

Abstract

Abstract-Recognition of Tibetan wood block print is a difficult problem that has many challenging steps. We propose a two stage framework involving image preprocessing, which consists of noise removal and baseline detection, and simultaneous character segmentation and recognition by the aid of a generalized hidden Markov model (also known as gHMM). For the latter stage, we train a gHMM and run the generalized Viterbi algorithm on our image to decode observations. There are two major motivations for using gHMM. First, it incorporates a language model into our recognition system which in turn enforces grammar and disambiguates classification errors caused by printing errors and image noise. Second, gHMM solves the segmentation challenge. Simply put gHMM is an HMM where the emission model allows multiple consecutive observations to be mapped to the same state. For features of our emission model we apply line and circle Hough transform to stroke detection, and use classspecific scaling for feature weighing. With gHMM, we find KMQDF to be the most effective distance metric for discriminating character classes. The accuracy of our system is 90.03%.

An Unsupervised Character Recognition Method for Tibetan Historical Document Images Based on Deep Learning

“…A method based on logical inference is proposed for slicing and recognizing characters in documents [8], and research is carried out on feature extraction [9] and recognition of similar characters in documents [10]. In 2011, Hedayati at the University of California, Berkeley, proposed a complete scheme for recognizing engraved Tibetan texts, addressing the baseline detection, line character slicing and various problems faced in the recognition process [11]. Since 2017, Wang Weilan's team from Northwest University for Nationalities and Duan Lijuan's team from Beijing University of Technology have been studying the recognition of Tibetan characters in Tibetan historical documents with the woodcut version of the Ugandan script Tibetan characters, and have made some research results in image pre-processing [12], layout analysis [13,14], text line slicing [15][16][17][18], character slicing [19] and data set construction [20] for the characters of Tibetan historical documents.…”

Character Detection and Segmentation of Historical Uchen Tibetan Documents in Complex Situations

“…From 1991, Kojima et al [1]- [3] studied the recognition of Tibetan documents in woodcuts. Since 2010, researchers have carried out relevant studies on image preprocessing [4] [5], layout analysis [6] [8], text line segmentation [9]- [12], character segmentation [13] [14], dataset construction [15] [16], character recognition [17] [18] and other aspects of historical Tibetan documents.…”