Abstract: The present invention is a method of optical character recognition. First, text is received. Next all words in the text are identified and associated with the appropriate line in the document. The directional derivative of the pixellation density function defining the text is then taken, and the highest value points for each word are identified from this equation. These highest value points are used to calculate a baseline for each word. A median anticipated baseline is also calculated and used to verify each baseline, which is corrected as necessary. Each word is then parsed into feature regions, and the features are identified through a series of complex analyses. After identifying the main features, outlying ornaments are identified and associated with appropriate features. The results are then compared to a database to identify the features and then displayed.