Segmentation of Touching Handwritten Digits Using Self-Organizing Maps

This paper presents a new algorithm for segmentation of touching handwritten digits. The proposal is divided into two parts: the selection of feature points which is made after the application of skeletonization, and the use of Self-Organizing Maps to define the segmentation points. As these two parts are independent from each other, the method is suitable for parallelization, increasing its performance. The algorithm was tested in a set of 42 images of real digits synthetically connected and achieved very promising results.