An integrated approach for automatic semantic structure extraction in document images

摘要

In this paper we present an integrated approach for semantic structure extraction in document images. Document images are initially processed to extract both their layout and logical structures on the base of geometrical and spatial information. Then, textual content of logical components is employed for automatic semantic labeling of layout structures. To support the whole process different machine learning techniques are applied. Experimental results on a set of biomedical multi-page documents are discussed and future directions are drawn.