Segmentation of single- or multiple-touching handwritten numeral string using background and foreground analysis

A new approach of segmenting single- or multiple-touching handwritten numer al string (two-digits) is proposed. Most algorithms for segmenting connecte d digits mainly focus on the analysis of foreground pixels. Some concentrat ed on the analysis of background pixels only and others are based on a reco gnizer, in this paper, we combine background and foreground analysis to seg ment single- or multiple-touching handwritten numeral strings. Thinning of both foreground and background regions are first processed on the image of connected numeral strings and the feature points on foreground and backgrou nd skeletons are extracted. Several possible segmentation paths are then co nstructed and useless strokes are removed. Finally, the parameters of geome tric properties of each possible segmentation paths are determined and thes e parameters are analyzed by the mixture Gaussian probability function to d ecide the best segmentation path or reject it. Experimental results on NIST special database 19 (an update of NIST special database 3) and some other images collected by ourselves show that our algorithm can get a correct rat e of 96 percent with rejection rate of 7.8 percent, which compares favorabl y with those reported in the literature.