Recognition of printed muti-font alphanumeric characters of multisize using 3×3 templates

We propose an algorithm which recognizes multi-font alphanumeric and special characters of multisize. The algorithm reduces the image size of each character read from the scanner using a 3×3 image array and normalizes a character to a 10×8 size for character recognition of multisize characters. After 4-directional features and ratios of binary 1s are obtained from the topology of a character, the character is changed into an integer in order to speed up the process. The algorithm recognizes the character among the candidate characters by template matching using the logical exclusive-OR proposed. In experimental results, the recognition ratio is about 98.3% and the recognition speed is about 14 characters per second for the 12224 characters of three different fonts and three different sizes