Character Recognition Of Multifont And Any Size For Medical Image Indexing System

This paper introduces a system that separates and recognizes text from images with format of DICOM, which is based on standard of digital imaging and communication in medicine. Considering that different kinds of images are saved together to be recognized, the system is able to recognize characters of various fonts and sizes. Method of feature extraction is presented, combining outer contour variation, geometric feature and black-hopping, etc. Experiments have shown better than 98 percent correct performance at a speed of 32 characters/s on mixtures of seven dissimilar fonts, over a range of point sizes.