Kinetic Online Trajectory Recovery from Static Images of Handwriting

Thumbnail Image
Issue Date
Journal Title
Journal ISSN
Volume Title
In this Bachelor thesis I propose two new approaches for extracting online handwriting data from scanned images of handwriting (offline data) using the mechanical concept of the momentum. The momentum describes the movement of a body and can only be changed by exerting forces, as first described by Sir I. Newton [1]. The first approach extracts pen strokes directly from an offline handwriting by tracing lines with a tracing point that uses a momentum. The second approach tries to sequence pen strokes to whole pen trajectories by using a kinetic cost function that is based on concepts derived from the definition of the momentum. An exploration of the limits of the first approach shows that it is not capable of dealing with noise that occurs in normal handwriting. The kinetic cost function of the second approach is compared to a traditional Euclidean distance based cost function for stroke sequencing. Using the kinetic cost function for stroke sequencing leads to significantly better pen trajectories than using the Euclidean distance cost function. Using a momentum based cost function for sequencing pen strokes can improve the quality of extracted pen trajectories.
Faculteit der Sociale Wetenschappen