Recovering Drawing Trajectory

Recovering Drawing Trajectory

Mehdi Iranpoor (Amirkabir University of Technology, Iran)
DOI: 10.4018/978-1-4666-2661-4.ch017


Graph theory has many applications in solving real-life problems. However, the application of Eulerian graphs and Eulerian tours/trails seems to be comparatively limited. In this chapter, an application of graph theory in handwriting recognition is presented. There are a lot of studies regarding handwriting recognition. Most of these methods deal with either offline or online handwriting recognition. However, the discussed approaches in this chapter are distinct in the manner that they aim to extract the trajectory of writing so as to equip the offline handwritten image with temporal information. When the trajectory of writing is known, it can be possible to utilize online recognition methods which are more reliable. These trajectory extracting methods are based on Eulerian trails in semi-Eulerian graphs. Semi-Eulerian graphs are graphs with at most two odd vertices. Eulerian trail is a trail in which every edge is traversed exactly once. The methods can be helpful in recognition of single-stroke handwritten images. Relying on the minimum energy law, the methods try to find the smoothest trajectory of writing which contribute to the recognition process.
Chapter Preview


Handwriting recognition is vital at least so long as the individuals write on papers or digital surfaces. In a comprehensive survey, Plamondon and Srihari reviewed many of online and offline approaches in handwriting recognition(Plamondon & Srihari, 2000).

Vinciarelli and Perrone proposed a combined method for recognition of online handwriting(Vinciarelli & Perrone, 2003). In their approach, first the online handwriting is recognized by an online system. Then, it is converted to a bitmap image and is recognized by an offline recognizer. By combining both results, the ultimate recognition becomes available which is more reliable than each individual result.

Trajectory recovery is an approach of extracting the chronology of writing of segments in order to enhance the reliability of offline handwriting recognition. An up-to-date review of trajectory extraction of offline handwriting approaches including analysis of different input types (skeleton or contour), end point detection, and trajectory recovery is presented by Nguyen and Blumenstein(Nguyen & Blumenstein, 2010). Nel uses Hidden Markov Models to estimate the trajectory of handwritten image(Nel, 2005). HMM is a probabilistic model which describes the dynamic transition between states. Her numerical results show that the method is both accurate and robust.

Complete Chapter List

Search this Book: