Enhancing handwritten character images thanks to a re-sampling process based on convex hull extraction

In this paper, we propose a new method that allows, by finding the convex hull of a character image, to set out in one pass only, the control parameters of a particular character distortion process. This character distortion method can then be applied to normalize the character image, i.e. to reduce the within-class scatter of images of handwritten characters, which could lead to a significant improvement of recognition performance. Many tests have been performed on unconstrained handwritten uppercase letters extracted from the NIST3 database [1]. Finally, the combination of two classifiers, one using the proposed normalization method, and the other one not, has allow reducing the overall error rate from 5.24% to 3.88%.