Caption text extraction for indexing purposes using a hierarchical region-based image model | Publicación