摘要 |
A method of recognizing a document image including a plurality of areas is provided. The method includes the steps of inputting the document image as a digital image, specifying a background color of the document image, reducing the size of the document image if necessary, extracting a plurality of pixels located in areas other than a background area from the document image by use of the background color, creating a plurality of connected elements by combining the plurality of pixels, and classifying the plurality of connected elements into a plurality of fixed types of areas by using at least features of shapes of the plurality of connected elements to obtain an area-separated document image. Additionally, a feedback process is performed if necessary, by performing binary area separation to a binary image and comparing a result of the binary area separation and a result of color area separation, to create a binary image suitable for an OCR process and to obtain a result of area separation.
|