Multi-font Numeral Recognition without Thinning based on Directional Density of Pixels

In this paper a fast and novel method is proposed for Multi-font numeral recognition which is a thinning free approach. The minimum rectangle Bounding box is fitted over isolated numeral image and image is cropped. The outer densities of pixels for each of the direction are computed in four directions viz. bottom, top, left and right. The ratios of these densities are taken with the total area of the cropped numeral image and are stored in a feature vector. The images are trained for the 16 different font styles by considering one image per numeral style and the mean feature set is estimated and stored as the library. A decision tree based minimum distance nearest neighbor classifier is used to classify English numerals by varying the size of numeral image between sizes 8 to 72 for the 16 different font styles. The total of 3200 numeral images tested and the overall accuracy of classification is found to be 99. 78 %. The average time taken by this method is 0.0160 seconds. The novelty of this method is that it is thinning-free, fast, accurate and showed encouraging results on multi-font and sizes.