Image analysis – Learning systems – Trainable classifiers or pattern recognizers
Patent
1995-04-28
1999-09-21
Chang, Jon
Image analysis
Learning systems
Trainable classifiers or pattern recognizers
345467, 382112, 382187, 382229, G06K 962
Patent
active
059564199
ABSTRACT:
A method for operating a machine to perform unsupervised training of a set of character templates uses as the source of training samples an image source of character images, called glyphs, that need not be manually or automatically segmented or isolated prior to training. A recognition operation performed on the image source of character images produces a labeled glyph position data structure that includes, for each glyph in the image source, a glyph image position in the image source associating an estimated image location of the glyph in the image source with a character label paired with the glyph image position that indicates the character in the character set being trained. The labeled glyph position data and the image source are then used to determine sample image regions in the image source; each sample image region is large enough to contain at least a single glyph but need not be restricted in size to only contain a single glyph. The template construction process using unsegmented samples is mathematically modeled as an optimization problem that optimizes a function that represents the set of character templates being trained as an ideal image to be reconstructed to match the input image. The method produces all of the character templates substantially contemporaneously by using a novel pixel scoring technique that implements an approximation of a maximum likelihood criterion subject to a constraint on the templates produced which holds that foreground pixels in adjacently positioned character images have substantially nonoverlapping foreground pixels. The character templates produced may be binary templates or arrays of probability values.
REFERENCES:
patent: 3233219 (1966-02-01), Atrubin et al.
patent: 4769716 (1988-09-01), Casey et al.
patent: 5020112 (1991-05-01), Chou
patent: 5303313 (1994-04-01), Mark et al.
patent: 5321773 (1994-06-01), Kopec et al.
patent: 5361356 (1994-11-01), Clark et al.
patent: 5526444 (1996-06-01), Kopec et al.
patent: 5706364 (1998-01-01), Kopec et al.
patent: 5828771 (1998-10-01), Bloomberg
Zhao et al., Cursivewriter: On-Line Cursive Writing Recognition System, Proceedings of the Second International Conference on Document Analysis and Recognition, IEEEComput. Soc. Press, pp. 703-706, Oct. 1993.
Kopec et al., Automatic Generation of Custom Document Image Decoders, Proceedings of the Second International Conference on Document Analysis and Recognition, IEEE Comput. Soc. Press, pp. 684-687, Oct. 1993.
National Science Foundation (NSF) Grant Proposal for NSF Program Digital Libraries NSF 93-141 Feb. 4, 1994, submitted by the Regents of the University of California, Berkeley, document date Jan. 24, 1994, pp. i-xi, 2-5, 36-37, 101, and 106.
G. Kopec, "Least-Squares Font Metric Estimation from Images", in IEEE Transactions on Image Processing, Oct., 1993, pp. 510-519.
G. Kopec and P. Chou, "Document Image Decoding Using Markov Source Models." in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 16, No. 6, Jun. 1994, pp. 602-617.
P.A. Chou, "Recognition of Equations Using a Two-Dimensional Stochastic Context-Free Grammar," in SPIE, vol. 1199, Visual Communications and Image Processing IV, 1989, pp. 852-863.
L. Rabiner, "A Tutorial on Hidden Markov Models and Selected Applications in Speech Recognition", in Proceedings of the IEEE, vol. 77, No. 2, Feb., 1989, at pp. 257-285.
H. S. Baird, "A Self-Correcting 100-Font Classifier," in SPIE vol. 2181 Document Recognition, 1994, pp. 106-115.
S. Kuo and O.E. Agazzi, "Keyword spotting in poorly printed documents using pseudo 2D hidden Markov models," in IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 16, No. 8, Aug., 1994, pp. 842-848.
C. Bose and S. Kuo, "Connected and degraded text recognition using a hidden Markov model," in Proceedings of the International Conference on Pattern Recognition, Netherlands, Sep. 1992, p. 116-119.
E. Levin and R. Pieraccini, "Dynamic planar warping for optical character recognition," in Proceedings of the 1992 International Conference on Acoustics, Speech and Signal Processing ("ICASSP"), San Francisco, California, Mar. 23-26, 1992, pp. III-149--III-152.
C. Yen and S. Kuo, "Degraded document recognition using pseudo 2D hidden Markov models in gray-scale Images". Copy received from authors without obligation of confidentiality, upon general request by applicants for information about ongoing or new work in this field. Applicants have no knowledge as to whether subject matter in this paper has been published.
R. Rubenstein, Digital Typography: An Introduction to Type and Composition for Computer System Design, Addison-Wesley, 1988, pp. 115-121.
Adobe Systems, Inc. Postscript Language Reference Manual, Addison-Wesley, 1985, pp. 95-96.
A. Kam and G. Kopec, "Separable source models for document image decoding", conference paper presented at IS&T/SPIE 1995 Intl. Symposium on Electronic Imaging, San Jose, CA, Feb. 5-10, 1995.
A. Kam, "Heuristic Document Image Decoding Using Separable Markov Models", S.M. Thesis, Massachusetts Institute of Technology, Cambridge, MA, Jun., 1993.
P.A. Chou and G.E. Kopec, "A Stochastic Attribute Grammar Model of Document Production and Its Use in Document Image Decoding", conference paper presented at IS&T/SPIE 1995 Intl. Symposium on Electronic Imaging, San Jose, CA, Feb. 5-10, 1995.
O. E. Agazzi et al., "Connected and Degraded Text Recognition Using Planar Hidden Markov Models," in Proceedings of International Conference on Acoustics, Speech and Signal Processing ("ICASSP") 1993, Apr. 1993, pp. V-113--V-116.
Chou Philip Andrew
Kopec Gary E.
Chang Jon
Patel Jayanti K.
Xerox Corporation
LandOfFree
Unsupervised training of character templates using unsegmented s does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Unsupervised training of character templates using unsegmented s, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Unsupervised training of character templates using unsegmented s will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-87683