Text/Non-text Classification in Online Handwritten Documents with Conditional Random Fields

In this work, we present a new method for discriminating textual from non-textual ink strokes in unconstrained handwritten online documents. A Conditional Random Field is utilized for jointly modeling several sources of information (local, spatial, temporal) that contribute to improve the classification accuracy at the stroke level. Experiments over the publicly available IAM-OnDo database validate the approach with an overall recognition rate of more than 96%, and highlight the contributions of the different sources of contextual information.

[1]  Marcus Liwicki,et al.  IAMonDo-database: an online handwritten document database with non-uniform contents , 2010, DAS '10.

[2]  Anil K. Jain,et al.  Structure in on-line documents , 2001, Proceedings of Sixth International Conference on Document Analysis and Recognition.

[3]  Christopher M. Bishop,et al.  Distinguishing text from graphics in on-line handwritten ink , 2004, Ninth International Workshop on Frontiers in Handwriting Recognition.

[4]  Masaki Nakagawa,et al.  Separating drawings, formula and text from free handwriting , 2003 .

[5]  David Jones,et al.  Discerning structure from freeform handwritten notes , 2003, Seventh International Conference on Document Analysis and Recognition, 2003. Proceedings..

[6]  Cheng-Lin Liu,et al.  Text/Non-text Ink Stroke Classification in Japanese Handwriting Based on Markov Random Fields , 2007 .

[7]  Andrew McCallum,et al.  Conditional Random Fields: Probabilistic Models for Segmenting and Labeling Sequence Data , 2001, ICML.

[8]  Jorge Nocedal,et al.  On the limited memory BFGS method for large scale optimization , 1989, Math. Program..