Graph Models and Maximum Common Subgraph for Character Analysis

A criticism of current image-to-graph transformations, and an investigation within simple domains.

An investigation into graph models within computer vision—the project provides a criticism of existing general techniques reliant upon the Delaunay triangulation. I also propose and investigate techniques for modelling and recognition of characters, against several well-known handwritten datasets as well as a sampling of fonts. The attached paper constitutes my Masters dissertation.