Image Extracting from Ancient Arab Documents with Complex Structures

In this paper, we give an overview on approaches for the graphic extraction, and especially for the image extraction from printed documents. We present also our contribution to this area operating on historical Arab periodicals. The developed method works on monochrome documents and is based on the Gabor filter exploration followed by many post-processing steps. It allows firstly the text/graphic separation then it ensures the distinction between images and other graphic classes (drawings, textured titles). The various tests and experiments were performed on an image database obtained from historical documents with complex structures, collected from various newspapers coming from the National Archives of Tunisia. The obtained results and the comparison carried out with another existing approach show many interesting perspectives and prove that our approach is able to attain high level of acceptability.