Recognition of printed multifont alphanumeric characters using Walsh transform function

In this paper, we propose an algorithm which recognizes multisize multi-font alphanumeric and special characters. The algorithm reduces the image size of each character read from the scanner into a 3×3 image array and normalizes each character to 8×8 size for character recognition. In other words, the segmented input characters are 1/9 by 3×3 of the image array, where the size of the image array can be determined from the size of the characters. After the 4-directional features are obtained from the Walsh transform function and the ratios of binary 1´s are obtained from the topology of the characters, each character is changed into an integer in order to reduce the memory capacity required, both memory space and recognition time is saved. In experimental studies, the recognition ratio is about 98.8% after postprocessing for the 14,120 characters of three different fonts