Cursive script segmentation incorporating knowledge of writing

In this paper a novel approach to character segmentation in cursive script is proposed based on knowledge of the writing. The knowledge used in the system includes the structural information of some prominent characters, as well as rules defining the particular combinations of background regions and character components. The segmentation is implemented in two steps. First, the connected components consisting of more than one character are split into sub-components using face-up and face-down background regions. Then, the over-split sub-components are merged into characters according to the knowledge of character structures and their combination characteristics. The experimental results are promising. The main merit of this approach is that it combines image processing techniques with heuristic rules, which are in the form of forbidden rules and feasible rules, to enhance performance of segmentation