Mending Broken Metal Label Characters Based on Structural Features

In the process of the recognition of metal label characters, the characters often have broken gaps, which bring troubles to the next process. In this paper, a novel method is developed which is based on the characters structural features and the previous statistic information, and this method bridges the gap effectively. As the experiments show, using this method the reconstructed characterpsilas skeleton is smooth and structure is complete, which meets the needs of the next feature extraction.