Recovering the recording sequence in scanned handwritten texts

A. Yu. Saparov

Vestnik Udmurtskogo universiteta. Matematika, mehanika, kompʹûternye nauki, Tome 28 (2018) no. 4, pp. 595-610

Résumé

The article deals with the problem of recognizing handwritten texts from raster images. A method to recover the sequence of records in a handwritten text is described, that will reduce the task of offline-recognition to the task of online-recognition. The method is based on finding the Eulerian path with the minimum weight in the handwritten symbol skeleton graph. Some numerical characteristics are considered as weights, they show the complexity of the transition from one edge to another through a common vertex. A table of all possible combinations of pairs is constructed for this purpose. If there isn't Eulerian path in the original graph, the path is searched with the minimum number of breaks. The definition of a virtual edge is introduced, the transition on it is the formation of a gap in the path. It is necessary to split edges into pairs and calculate the weights at the vertices of odd multiplicity. The pathfinding algorithm in the skeleton of a symbol is considered, it is based on the Fleury's algorithm of searching Eulerian path.

Détail

Keywords: graph of a handwritten symbol skeleton, path in the skeleton, virtual edge.

Bibliographie

[1] Abramenko A., Principles of recognition, Computer-press, K., 1997, 123 pp.

[2] Kuchuganov A. V., Lapinskaya G. V., “Recognition of handwritten texts”, Modern information technologies and written heritage: from ancient manuscripts to electronic texts, Proceedings of the International Scientific Conference, Izhevsk State Technical University, Izhevsk, 2006, 98–103 (in Russian)

[3] Omar M., Omar F., Ismoilov M. I., Ostroukh A. V., “Using of pattern recognition in various specialization”, Automation and Control in Technical Systems, 2014, no. 4, 32–47 (in Russian) | DOI

[4] Katasev A. S., Kataseva D. V., Kirpichnikov A. P., “Handwritten character recognition based on artificial neural network”, Vestnik Kazanskogo Tekhnologicheskogo Universiteta, 18:11 (2015), 173–176 (in Russian)

[5] Jaderberg M., Simonyan K., Vedaldi A., Zisserman A., “Reading text in the wild with convolutional neural networks”, International Journal of Computer Vision, 116:1 (2015), 1–20 | DOI | MR

[6] Favorskaya M. N., Goroshkin A. N., “The invariant model for image recognition of hand-written text”, Vestnik Sibirskogo Gosudarstvennogo Aerokosmicheskogo Universiteta Imeni Akademika M. F. Reshetneva (Vestnik SibGAU), 2008, no. 2 (19), 52–57 (in Russian)

[7] Khaustov P. A., “Algorithms for handwritten character recognition based on constructing structural models”, Computer Optics, 41:1 (2017), 67–78 (in Russian) | DOI

[8] Zapryagaev S. A., Sorokin A. I., “Handwritten character recognition based on analysis of chord length function descriptors”, Vestnik Voronezhskogo Gosudarstvennogo Universiteta. Ser. Sistemnyi Analis i Informatsionnye Tekhnologii, 2009, no. 2, 49–58 (in Russian)

[9] Spitsyn V. G., Bolotova Yu. A., Phan N. H., Bui T. T. T., “Using a Haar wavelet transform, principal component analysis and neural networks for OCR in the presence of impulse noise”, Computer Optics, 40:2 (2016), 249–257 (in Russian) | DOI

[10] Ray A., Rajeswar S., Chaudhury S., “A hypothesize-and-verify framework for text recognition using deep recurrent neural networks”, 2015 13th International Conference on Document Analysis and Recognition (ICDAR) (Tunis, 2015), 936–940 | DOI

[11] Ray A., Rajeswar S., Chaudhury S., “OCR for bilingual documents using language modeling”, 2015 13th International Conference on Document Analysis and Recognition (ICDAR) (Tunis, 2015), 1256–1260 | DOI

[12] Tapia E., Understanding mathematics: a system for the recognition of on-line handwritten mathematical expressions, Dissertation, Berlin, 2004, 109 pp. https://pdfs.semanticscholar.org/2246/4fae30aa28706c323a0930ee82cb7f74856b.pdf

[13] Kosmala A., Rigoll G., “On-line handwritten formula recognition using statistical methods”, Proceedings. Fourteenth International Conference on Pattern Recognition (Brisbane, Australia, 1998), 1306–1308 | DOI

[14] Kosmala A., Rigoll G., Brakensiek A., “Online handwritten formula recognition with integrated correction recognition and execution”, Proceedings 15th International Conference on Pattern Recognition (Barcelona, Spain, 2000), v. 2, Pattern Recognition and Neural Networks, 590–593 | DOI

[15] Toyozumi K., Suzuki T., Mori K., Suenaga Y., “A system for real-time recognition of handwritten mathematical formulas”, Proceedings of Sixth International Conference on Document Analysis and Recognition (Seattle, WA, USA, 2001), 1059–1063 | DOI

[16] Gonzalez R. C., Woods R. E., Digital image processing, Prentice-Hall, 2007, 976 pp.

[17] Christofides N., Graph theory. An algorithmic approach, Academic Press, 1975, 400 pp. | MR | Zbl

[18] Saparov A. Yu., Bel'tyukov A. P., “Mathematical modeling of formula images for their recognition”, Vestnik Udmurtskogo Universiteta. Matematika. Mekhanika. Komp'yuternye Nauki, 2013, no. 1, 153–167 (in Russian) | DOI

[19] Knyazev A. V., “Recognition of joined-up handwritten words by means of fuzzy glyraphs”, Vestnik MEI, 2017, no. 5, 117–120 (in Russian) | DOI

[20] Isupov N. S., Kuchuganov A. V., “Graph theory application in handwriting recognition task,”, Vestnik Izhevskogo Gosudarstvennogo Tekhnicheskogo Universiteta Imeni M. T. Kalashnikova, 2012, no. 4, 160–162 (in Russian)

[21] Saparov A. Yu., “Recovering the recording sequence in scanned handwritten texts”, The 10th All-Russian Multiconference on Management Problems: Multiconference materials, v. 1, South Federal University, Rostov-on-don, 2017, 93–95 (in Russian)