Cursive word recognition using a random field based hidden Markov model

Abstract. In this paper we present a two-dimensional stochastic method for the recognition of unconstrained handwritten words in a small lexicon. The method is based on an efficient combination of hidden Markov models (hmms) and causal Markov random fields (mrfs). It operates in a holistic manner, at the pixel level, on scaled binary word images which are assumed to be random field realizations. The state-related random fields act as smooth local estimators of specific writing strokes by merging conditional pixel probabilities along the columns of the image. The hmm component of our model provides an optimal switching mechanism between sets of mrf distributions in order to dynamically adapt to the features encountered during the left-to-right image scan. Experiments performed on a French omni-scriptor, omni-bank database of handwritten legal check amounts provided by the A2iA company are described in great detail.

[1]  George Saon,et al.  Binary pattern recognition using Markov random fields and HMMs , 1997, 1997 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[2]  Seong-Whan Lee,et al.  An HMMRF-based statistical approach for off-line handwritten character recognition , 1996, Proceedings of 13th International Conference on Pattern Recognition.

[3]  H. Derin,et al.  Discrete-index Markov-type random processes , 1989, Proc. IEEE.

[4]  Laveen N. Kanal,et al.  Classification of binary random patterns , 1965, IEEE Trans. Inf. Theory.

[5]  George Saon,et al.  High Performance Unconstrained Word Recognition System Combining HMMs and Markov Random Fields , 1997, Int. J. Pattern Recognit. Artif. Intell..

[6]  Oscar E. Agazzi,et al.  Keyword Spotting in Poorly Printed Documents using Pseudo 2-D Hidden Markov Models , 1994, IEEE Trans. Pattern Anal. Mach. Intell..

[7]  G. Saon,et al.  Off-line handwritten word recognition using a mixed HMM-MRF approach , 1997, Proceedings of the Fourth International Conference on Document Analysis and Recognition.

[8]  Pierre A. Devijver,et al.  Learning the parameters of a hidden Markov random field image model: A simple example , 1987 .

[9]  Fumitaka Kimura,et al.  Improvements of a lexicon directed algorithm for recognition of unconstrained handwritten words , 1993, Proceedings of 2nd International Conference on Document Analysis and Recognition (ICDAR '93).

[10]  Pierre A. Devijver,et al.  Image Segmentation Using Causal Markov Random Field Models , 1988, Pattern Recognition.

[11]  John W. Woods,et al.  On the relationship of the Markov mesh to the NSHP Markov chain , 1987, Pattern Recognit. Lett..

[12]  Nikolai Gorski,et al.  The A2iA Intercheque System: Courtesy Amount and Legal Amount Recognition for French Checks , 1997, Int. J. Pattern Recognit. Artif. Intell..

[13]  Gilloux 04 - Reconnaissance de chiffres manuscrits par modèle de Markov pseudo-2D , 1995 .

[14]  D. Rubin,et al.  Maximum likelihood from incomplete data via the EM - algorithm plus discussions on the paper , 1977 .

[15]  Roberto Pieraccini,et al.  Dynamic planar warping for optical character recognition , 1992, [Proceedings] ICASSP-92: 1992 IEEE International Conference on Acoustics, Speech, and Signal Processing.

[16]  Oscar E. Agazzi,et al.  Hidden markov model based optical character recognition in the presence of deterministic transformations , 1993, Pattern Recognit..

[17]  Patrick Bouthemy,et al.  Multimodal motion estimation and segmentation using Markov random fields , 1990, [1990] Proceedings. 10th International Conference on Pattern Recognition.