Skip to main content.

INK DIVIDER

Inked documents, for instance a sketched diagram, annotations on an article or to-do list, generally consist of both writing and drawing elements. From a recognition perspective the writing and drawing are fundamentally different. Writing strokes are joined together to create letters, words, phrases and sentences: and the writing has an abstract meaning that is not directly related to the ink shapes. In contrast drawing ink is generally in the form of ideograms or pictograms where the meaning is much more directly related to the ink shapes. Because of this fundamental difference it is important to be able to accurately separate writing and drawing ink. This is a very challenging problem. So difficult that in fact most sketch tools do not attempt it, resorting to keyboard input of words or very constrained writing input. Clear, well researched methods of separating writing and drawing would be of great benefit for ink recognition and the pen-input research community and enable effective electronic pen-based solutions for a much wider range of domains.

We propose the use of data mining techniques to build more accurate text-shape dividers. This systematic approach identifies the algorithms best suited to the specific problem. We have generated dividers using data mining with diagrams from three domains and a comprehensive ink feature library and have found these to be more accurate than three existing dividers.

Text-Shape Division
Text-Shape Division Process

Publications

  • Blagojevic R., B. Plimmer, et al. (2010). Building Digital Ink Recognizers using Data Mining: Distinguishing Between Text and Shapes in Hand Drawn Diagrams, Proceedings of IEA-AIE 2010.
  • Blagojevic R., B. Plimmer, et al. (2008). Development of Techniques for Sketched Diagram Recognition, Proceedings of IEEE Symposium on Visual Languages and Human-Centric Computing Graduate Consortium, 258 - 259.
  • Patel, R., B. Plimmer, et al. (2007). Ink Features for Diagram Recognition. 4th Eurographics Conference on Sketch Based Interfaces and Modelling - Riverside, USA.

    • Other Publications

    This project was partially funded by Microsoft® Research Asia.