| 摘要 |
PURPOSE:To enable detection and elimination even when linear elements such as scoring or character frames are included in character string areas cut-out by dividing the document image into small areas with directionability and using the picture element feature extraction in the respective small areas. CONSTITUTION:The document image 23 stored by the document input memory means 16 is divided crosswise and small areas are prepared. Next, the black picture element in each small area is counted and when the value is >=5 the threshold value, it is binarized as '1', and the others as '0', and with each small area corresponding to 1 picture element, concentration feature image 24 is obtained. On the other hand, each small area is binarized as '1' when the black picture element from extreme right to left is connected with 8 links and as '0' for the others, and each small area corresponding to 1 picture element, linear feature image 27 is obtained. Next, by the linear element existence/ non-existence judgment means 20, where the value in the area of linear feature image 27, proportionate to 1 picture element is judged as existing in the character string area image 30. Thereby, linear element elimination means 21 is executed and the linear element pattern is eliminated, giving character string area image 31.
|
