Document layout analysis is a tough task for a document analysis and recognition system, especially when there are many variations in the layouts. Often, layout analysis requires character recognition, while character recognition requires layout analysis beforehand. It's a Catch-22, or a “chicken-and-egg” problem. This chapter discusses this kind of dilemma and presents a two-part solution that first analyzes the layout and then, using a hypothesis-driven approach, segments the numerical character line. Basically, the approach is to first generate multiple hypotheses based on low-level image processing, then to conduct layout analysis and create many candidates based on each of the hypotheses. Finally, the correct candidate is selected by the results of content recognition. Probabilistic verification is used to select the candidates that are input to the recognition module, with parameters that are learned from samples in advance. The second part of the solution, which relies on a hypothesis-driven approach for the segmentation of the numerical character line will also be presented. As a test case, these solutions were applied to the Japanese postal address recognition system. They show how tough problems involved in analyzing the surface images of mail pieces can be solved. The hypotheses-driven approach manages every possible variation, including writing orientation, printed or hand-written, address-block location, character size, and so on.
Access provided by Autonomous University of Puebla. Download to read the full chapter text
Chapter PDF
Similar content being viewed by others
Keywords
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
References
. Kobayashi, Y., Yamada, K., Tsukumo, J.: A Segmentation Method for Hand-Written Japanese Character Lines Based on Transitional Information. Proc. ICPR (1992) 487-491
. Kamimura, T., et al.: Determing Address Format Using Layout Rules for Read-ing Japanese Hand-written Mail. (in Japanese) IEICE PRU95-107 (1995) 25-30
Yeh, P.S., et al.: Address Location on Envelopes. Pattern Recognition (1987) Vol. 20, No. 2, 213-227
. Wolf, M., et al.: Form-Based Localization of the Destination Address Block on Complex Envelopes. Proc. ICDAR (1997) 908-913
. Wolf, M., et al.: Fast Address Block Location on Handwritten and Machine Printed Mail-piece Images. Proc. ICDAR (1997) 753-757
. Yu, B., et al.: Address Block Location on Complex Mail Pieces. Proc. ICDAR (1997) 897-901
. Sako, H., et al.: On The Rejection Ability Required in Multiple Hypothesis Techniques. Proc. IWFHR (2000) 123-132
Casey, R.G., et al.: A Survey of Methods and Strategies in Character Segmen-tation. IEEE Trans. on PAMI (1996) Vol. 18, No. 7, 690-706
Fujisawa, H., et al.: Segmentation Method for Character Recognition. Proc. IEEE (1992) Vol. 80 No. 7, 1079-1092
. Gyohten, K., Sumiya, T., Babaguchi, N., Kakusho, K., Kitahashi, T.: A Multi-Agent Based Method for Extracting Characters and Character Strings. IEICE TRNS. INF. & SYST. (1996) VOL. E79-D, NO. 5
. Bruckner, T., Suda, P., Block, H.U., Maderlechner, G.: In-house Mail Distribu-tion by Automatic Address and Content Interpretation. SDAIR96 67-75
. Srihari, S.N., Srikantan, G., Hong, T., Grom, B.: A General-Prupose Japanse Optical Character Recognition System. Proc. Conference on Document Recog-nition (1996) SPIE Symposium
Cohen, E., Hull, J.J., Srihari, S.: Control Structure for Interpreting Handwritten Addresses. IEEE TRANS. PAMI. (1994) Vol. 16, No. 10, 1049-1055
. Shinjo, H., Nakashima, K., Koga, M., Marukawa, K., Shima, Y., Hadano, E.: A Method for Connecting Disappeared Junction Patterns on Frame Lines in Form Documents. Proc. ICDAR (1997) 667-670
. Dunn, C.E., Wang, P.S.P.: Character Segmentation Techniques for Handwritten Text - A Survey. Pattern Recognition Methodology and Systems, Proceedings., 11th IAPR International Conference (1992) 577-580
. Tnag, Y.Y., Ma, H., Liu, J., Li, B.F., Xi, D.: Multiresolution Analysis in Ex-traction of Reference Lines from Documents with Gray Level Background. IEEE TRANS. PAMI. (1997) VOL. 19, NO. 8
. Koga, M., Kagehiro, T., Sako, H., Fujisawa, H.: Segmentation of Japanese Hand- written Characters Using Perioheral Feature Analysis. Proc. ICPR (1998) 1137-1141
. Sawaki, M.: Text-Line Extraction and Character Recognition of Document Headlines With Graphical Designs Using Complementary Similarity Measure. IEEE TRANS. PAMI. (1998) VOL. 20, NO. 10
Yen, P., Antoy, S., Litcher, A., Rosenfeld, A.: Address Location on Envelopes. Pattern Recognitoin (1987) Vol. 20, No. 2, 213-227
. Srihari, S.N.: Integration of Hand-Written Address Interpretation Technology into the United States Postal Service Remote Computer Reader System. Proc. ICDAR (1997) 892-896
. Palumbo, P.W., Srihari, S.N., Soh, J., Sridhar, R., Demjanenko, V.: Postal Ad- dress Block Location in Real Time. Computer (1992) 34-42
.Whichello, A.P., Yan, H.: Locating Address Blocks and Postcodes in Mail-Piece Images. Proc. ICPR (1996) 716-720
. Nakajima, N., Tsuchiya, T., Kamimura, T., Yamada, K.: Analysis of Address Layout on Japanese Handwritten Mail. Proc. ICPR (1996) 726-731
Liu, C., Koga, M., Fujisawa, H.: Gabor Feature Extraction for Character Recog-nition: Comparison with Gradient Feature. Proc. ICDAR (2005) Vol. 1, 121-125
Koga, M., Mine, R., Kameyama, T., Takahashi, T., Yamazaki, M., Yamaguchi, T.: Camera-based Kanji OCR for Mobile-phones: Practical Issues. Proc. ICDAR (2005) Vol. 2 635-639
Liu, C., Mine, R., Koga, M.: Building Compact Classifier for Large Character Set Recognition Using Discriminative Feature Extraction. Proc. ICDAR (2005) Vol. 2 846-850
Liu, C., Sako, H.: Class-specific Feature Polynomial Classifier for Pattern Clas-sification and its Application to Handwritten Numeral Recognition. Pattern Recognition (2006) Vol. 39 669-681
Kagehiro, T., Koga, M., Sako, H., Fujisawa, H.,: Address-Block Extraction by Bayesian Rule. Proc. ICPR (2004) Vol. 2 582-585
Ikeda, H., Furukawa, N., Koga, M., Sako, H., Fujisawa, H.: A Context-free Grammar-based Language Model for String Recognition. Int. J. Comput. Proc. Oriental Lang. (2002) Vol. 15, No. 2 149-163
. Shinjo, H., Hadano, E., Marukawa, K., Shima, Y., Sako, H.: A Recursive Anal-ysis for Form Cell Recognition. Proc. ICDAR (2001) 694-698
. Kagehiro, T., Koga, M., Sako, H., Fujisawa, H.: Segmentation of Handwritten Kanji Numerals Integrating Peripheral Information by Bayesian Rule. Proc. MVA (1998) 439-442
Author information
Authors and Affiliations
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2008 Springer-Verlag Berlin Heidelberg
About this chapter
Cite this chapter
Kagehiro, T., Fujisawa, H. (2008). Multiple Hypotheses Document Analysis. In: Marinai, S., Fujisawa, H. (eds) Machine Learning in Document Analysis and Recognition. Studies in Computational Intelligence, vol 90. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-540-76280-5_11
Download citation
DOI: https://doi.org/10.1007/978-3-540-76280-5_11
Publisher Name: Springer, Berlin, Heidelberg
Print ISBN: 978-3-540-76279-9
Online ISBN: 978-3-540-76280-5
eBook Packages: EngineeringEngineering (R0)