| 摘要 |
PROBLEM TO BE SOLVED: To reduce the extraction leakage of components even in a document picture consisting of document components having small color difference and a background by generating plural multivalued planes having different characteristics from a color document picture and extracting the elements of a document picture from the planes. SOLUTION: First, a document picture is produced by an image inputting means 101. Next, n pieces of plane generating means 102 generate plural image planes from the document picture. The planes to be generated are ones that represent characteristic values representing the R, G and B components of pixel value, luminance components, etc., with binary or multivalued parts and n pieces of planes are produced in this way. Then, a document element extracting means 103 extracts document elements such as characters, a background, a diagram and a ruled line from all of the n pieces of planes. A document recognizing means 104 performs character recognition of the obtained characters. Then, a document recognition integrating means 105 integrates character recognition results obtained from the n pieces of planes.
|
