K.K. Kim, J.Y. Lee, and J.H. Kim (Korea)
Camera document image, character segmentation
This paper presents character segmentation of camera document image that is based on merge-split method using structural information of character. Character segmentation is processed on binary image that includes extraction of character string and word, and individual character segmentation. To get clean binary image without background and to reduce illumination effect, local binarization method is processed. Projections and space information are used to extract character string and word. To segment individual character, block information is used that is defined using character structural information. Reliability factor using recognition probability is used to verify character segmentation. To evaluate the proposed segmentation method, we have experimented with camera document image of the ETRI database. An encouraging segmentation result and 92.5% of recognition result have been obtained.
Important Links:
Go Back
