| 摘要 |
PROBLEM TO BE SOLVED: To provide a method and device, extracting metadata from a document without requiring that the metadata is included inside a box structure or an apparently defined area delimiter. SOLUTION: In this method and device for extracting the metadata such as a title or an author from a document image (13) comprising pixels, at least one of the images is displayed on a display (12), to a user. A user's pointing control element such as a mouse or a touch screen is operated by the user, and a selection command is generated. The selection command includes a selection point among metadata elements of the image. An area comprising foreground pixels is determined. The area includes a pixel connected to the selection point. An extraction area (14) is constructed around the area. Lastly, the pixels of the extraction area are processed to extract the metadata. COPYRIGHT: (C)2005,JPO&NCIPI
|
