A Touching Pattern-Oriented Strategy for Handwritten Digits Segmentation

Among several contour splitting algorithms drop-fall algorithm is convenient and accurate for separating handwritten numerals. A set of drop-fall algorithms can produce four possible cutting paths. Based on four drop-fall algorithms, this paper deals with a pattern-oriented strategy for segmenting touching digits in order to solve the problem of determining which drop-fall algorithm producing the best separation. In this strategy, background region of touching digits is analyzed by using water reservoirs concept at first, and then the classification scheme is developed to identify the pattern of touching digits according to the structural features extracted from the specified background region. To split touching digits, the pattern-oriented strategy is implemented by choosing one suitable drop-fall algorithm applied to the pattern of current touching digits in terms of its performance characteristic. Experimental results show that the proposed strategy is effective for segmentation of handwritten touching digits.