A Junction Based Segmentation Algorithm for Offline Handwritten Connected Character Segmentation

A junction based approach for segmenting and recognizing offline handwritten connected two-digit strings is presented in this paper. Very often even in a printed text, adjacent characters tend to touch or connect. This makes it a problem in performing proper character isolation, hence difficult in segmenting the digit strings in order to recognize its individual characters. We, in our study have developed an algorithm, which provides a solution based on the analysis of the foreground pixel distribution to segment, connected digit string pairs. Each of pixels at the connected character skeleton is used to identify a major feature point called a junction point; a pixel point having three or more neighboring pixels, and possible segmentation paths are defined upon the junction points. In the segmentation stage, the junction based splitting technique decides complete segments of the connected digit strings. Use of fuzzy characteristic values at the merging of the complete segments isolates the major segments (Merged complete segments) from the minor segments. In this work, it was found that the unwanted connection resides within the minor segments of the connected character skeleton. At the character isolation stage, all major segments are combined with each of minor segments to generate set of different connection sketches. In order to recognize individual characters of the connection string, these sketches are formulated into a new input character image, which then can be used as an input for a character recognition system.