Modeling segmentation cuts using support vector machines

In this paper we propose a method to evaluate segmentation cuts for handwritten touching digits. The idea of this method is to work as a filter in segmentation-based recognition systems. These types of systems usually rely on over-segmentation methods, where several segmentation hypotheses are created for each touching group of digits and then assessed by a general-purpose classifier. Through the use of the proposed method, unnecessary segmentation cuts can be identified without any attempt of classification by a general-purpose classifier, reducing the number of paths in a segmentation graph, what can consequently lead to a reduction in computational cost. Concavity analysis is performed in each digit before and after segmentation. The difference of those concavities is used to model the segmentation cuts. SVM is used to classify those segmentation cuts. The preliminary results obtained are very promising as for the segmentation algorithm tested, 67.9% of the unnecessary segmentation cuts were eliminated. Moreover, it was possible to achieve a significant increase in the recognition rate for the generalpurpose classifier.