| 摘要 |
PURPOSE:To obtain a device which can correctly extract a discontinuous ruled line even when characters are close in a cell included in a table and correctly recognize table construction by providing a discontinuous line extracting part, etc., which extracts the ruled line excepting for a discontinuous line by investigating the posision relation of circumscribed data of which width/height are less than a threshold. CONSTITUTION:A labeling part 10 which selects a connecting black picture element from image data for table recognition to extract a label being a rectangular circumscribed by it and sets the lengths in a vertical/horizontal direction to be labeling data. A noize erasing part 11 fills the inside of the label of which vertical/horizontal lengths of labeling data is shorter than a noize threshold with white picture elements. A continuous ruled line extracting part 12 extracts a continuous ruled line by connecting connective ones among a line extracted by a continuous line extracting part 3. The discontinuous ruled line extracting part 13 extracts the discontinuous ruled line by investigating the position relation of the labels of which vertical/horizontal length of labeling data is shorter than the discontinuous ruled line width threshold. |
