Region and Keyword Extraction Based on Color Marking for Document Entry

A method for extracting an article and its keywords from a notated document using color image processing is presented. In this method, the user selects a region of an article and its keywords and marks them in color on the printed document. The system extracts the article and recognizes the keywords. This paper describes, (1) colored area extraction and color effect removal by adaptive thresholding using pseudoHVC coordinates based on the "transparent coloration model" which based on the behavior of colored pixels in color space, (2) polygonal shape extraction which extracts the intended article from a handdrawn outline by a new and simple algorithm "recursive vertex search", (3) and some experimental results which show high accuracy for newspaper article extraction using colored pixels in color space. The color marks of article borders and keywords are identified and then removed by adaptive thresholding using the pseudo-HVC coordinates. Because the color marks are handdrawn they will not be completely accurate. We have developed a polygonal shape extraction algorithm that examines the marked outline and then determines which part of the text is to be extracted. This algorithm is not yet complete because it needs information on typical document architecture. When complete the system will extract the desired article and stored as an image in an electronic filing system.