An Optimization Methodology for Document Structure Extraction on Latin Character Documents
IEEE Transactions on Pattern Analysis and Machine Intelligence
A new pyramidal approach for the address block location based on hierarchical graph coloring
ICIAR'07 Proceedings of the 4th international conference on Image Analysis and Recognition
| Hi-index | 0.00 |
This paper presents a page segmentation method called block selection which not only segments the page image into categorized blocks but also provides a novel tree structure to represent the page blocks for selection. Block selection, more than classifying the text and nontext areas only, can identify the major document elements, such as text, picture, table, frame and line. This ability fits block selection into a wider range of document processing applications. In order to make the usage of block selection more practical to various document styles, many restrictions set on the document by some existing technologies are freed. The language on the document could be English-like, Kanji-like or both. The direction of text could be horizontal, vertical, slanted, or mixed. The editing style of the document is unconstrained. No skew correction is involved regardless of the document style. The formed blocks are described by a hierarchical tree to reflect the page arrangement in the "object" sense. This structural result can be efficiently used for further storage, retrieval or other manipulation purposes. The possible applications using this proposed method are discussed.