Block Detection

Detection of (textual) blocks logically precedes text line detection and handwritten text recognition. Technically, textual blocks are represented as closed paths with four handlers at the "corners", which can be graphically adjusted by the user.

In GiDoc , block detection has been for now implemented to simply copy the blocks of the preceding image to the current image. This works quite well in the case of document collections with homogeneous page layout structure. Nevertheless, if the user is not interested in copying the blocks of the preceding image, then manual block annotation can be easily carried out by first selecting the appropriate "rectangular" area (with the Rectangular Selection tool), and then calling Block Detection. An illustrative example is shown in Fig. 6.

Figure 6: Example of a text block detected using the Block Detection tool.
Image layout0



giDoc Team