Segmentation of Touching Numeric Digits

Abstract

In this work, we have discussed the segmentation system for numeric digits. We have studied the existing algorithms used for the segmentation of numeric digits and tried to explain it. As by casey.et.al, there are three approaches for segmentation (classical, recognition based and holistic methods). But under present study we focused on the classical approach of segmentation for numeric digits, which is based on character features alone (i.e. not aided by recognition). Various techniques like Min-Max, Hit and Deflect, Structural feature based technique and drop fall have studied for numeric digit segmentation. The problem of segmenting overlapping or touching character has been studied in detail. Variations of drop fall have also discussed which have advantages over one another. A hybrid heuristic, which is the mixing of all these variations have discussed in detail, which has promising results. This algorithm is used by many OCR designers to have complete OCR system. An improved segmentation algorithm, Extended Drop fall has also been discussed for proper segmentation of problematic cases.