Based on the study of the specificity of historical printed books and on the main error sources of classical methods of page layout analysis, this paper presents a new way to achieve an indexation of ancient printed documents. We have developed an approach based on the extraction and the quantification of the various orientations that are present in printed document images. The documents are initially splitted into homogenous areas in which we analyze significant orientations with a directional rose. Each kind of information (textual or graphical) is typically identified and labelled according to its orientation distribution. This choice of characterization allows us to separate textual regions from graphical ones by minimizing the a priori knowledge. The evaluation of our proposition lies on a document image retrieval using layout extraction criteria and can also be used to precisely localize graphical parts in various types of documents. The system has been tested with success over several ancient printed books of the Renaissance.
展开▼
