| 摘要 |
A language such as English or a language group such as an Asian language group is recognized based upon document image data. The document image data is processed to determine a minimal circumscribing rectangle for each character. The layout characteristics of the minimal circumscribing rectangles are quantified in a discrete number of ranges. The layout characteristic information includes a certain ratio with respect to the minimal circumscribing rectangle height and width as well as a black pixel density in the minimal circumscribing rectangle. Based upon the quantified layout characteristic information, an occurrence probability of a predetermined number of characters is determined using training data for a predetermined number of languages. The occurrence probability is stored in a table for later reference for an unknown input language.
|
