A new method for the segmentation of mixed handprintedChinese/English characters

摘要

Describes a new method to segment mixed handprintedChinese/English characters. First, connected component analysis isperformed on each text line. A pre-merging step combines thosevertical-neighboring bounding rectangles into a new one, but leavesalone the horizontal-neighboring ones which may belong to differentcharacters. Then the feature-complexity analysis of each boundingrectangle is used to classify it into an alphanumeric character, anisolated Chinese radical or a complete Chinese character. Finally, basedon the recognition result of an isolated Chinese radical and heuristicformation rules, an isolated Chinese radical may be further combinedwith adjacent bounded rectangle(s) to make one single characterrectangle for a separate character. Implementation and experimentalresults of the proposed method are described