摘要 |
The device for coding/decoding binary document images employs a module 10 for extracting shapes defined as sets of connected points through recursively updated state vectors, a shape-recognition module 20 comparing the state vectors of the extracted shapes with the state vectors of the shapes previously encountered and stored in a library 30, performing a prerecognition by systematically comparing the extracted shape with all the library symbols, selecting a sub-set of near symbols, then a fine recognition by comparing the extracted shape with the near symbols, and a coding module 40 transmitting, after variable-length coding, for each shape extracted, the label of the identified library symbol and its position in the page, or a description of the unrecognised extracted shape and its position in the page. This device is applicable to fast, high-quality digital facsimile systems for the majority of normal documents. <IMAGE>
|