The representation process in both statistical and syntactical approaches is analyzed and discussed. A representation approach that incorporates the topological information about the image without the requirement for a pattern grammar and parser is introduced. This approach is believed to be applicable to a wide domain of situations. Results of implementing this approach for recognition of digitized English characters are given. The test data used for the experiments were obtained by digitizing English text at a density of 250 pixels/in. A sample digitized character is shown.
展开▼