In the framework of structural character recognition, the authorspresent a method to reduce the number of prototypes to match with agiven sample. The basic idea is that a coarse description of the sample,even if not adequate for the recognition, can be powerful enough todiscriminate among the prototypes those that most likely will match thesample. Once this subset has been found, a more detailed description iscomputed, and the main classification step entered. To achieve thepurpose, a multilevel description of the character, in terms of thefeatures provided by the feature extractor. At the intermediate level,the character is decomposed into components by removing the branchpoints. Eventually, each component is further split into simple,meaningful parts called superfeatures. By using the highest level of thedescription a fast and reliable selection of the prototypes to beconsidered as candidates for the matching can be obtained, while thelowest one is used by the main classifier to choose which one of theprototypes, among the selected ones, has the best matching with thesample. Experiments have proved that the method is correct andefficient. It is correct since it makes it possible to select a subsetof prototypes which always contains the right one, and it is efficientsince it significantly reduces the number of prototypes to be matchedwith the sample
展开▼
