Gabor filter based multi-class classifier for scanned document images

When scanning documents with a large number of pages such as books, it is often feasible to provide a minimal number of training samples to personalize the system to compensate for global shifts in how the document was created or in scanning parameters. In this paper, we present a supervised multi-class classifier based on Gabor filters that is used to classes the scripts, font faces, and font-styles (bold, italic, normal etc.) in an application where the classes are known. Classification is performed at the word level (glyphs separated by white space) given training samples of each class. This method was applied to a variety of bilingual dictionaries to identify different scripts, and simultaneously, to class Roman scripts into bold, italic and normal font-styles. Experimental results show the effectiveness of this approach in increasing performance over classifiers trained for general documents.