Unsupervised clustering of text entities in heterogeneous grey level documents

Presents a new method of functional classification of text blocks on a document. It is based on texture analysis and unsupervised classification. Texture is used to define different classes of text blocks in the document and to direct a possible way of exploration from the most eye-catching data to the less significant text block. The typographical properties of blocks are characterized by two main discriminating primitives : the complexity of the text drawing and the structural relief of the block. This analysis is the starting point of a three-classes categorization into functional families (main headings, sub-headings and text paragraphs). Each block of text is described and classified through a labeling process based on a 3D-feature space using the two previous features (complexity and structural relief) and a third one among pattern primitives, blocks size and location in the document. This method allows a first approach to a global context free classification of documents.