Visual language processing (VLP) of ancient manuscripts: Converting collections to windows on the past

摘要: Ancient manuscripts constitute a primary carrier of cultural heritage globally, and they are currently being intensively digitized all over the world to ensure their preservation, and, ultimately, wide accessibility content. Critical this research process legibility documents in image form, access live texts. Several state-of-the-art methods approaches have been proposed developed address challenges associated with processing these manuscripts. However, there is huge amount data involved, also high cost scarcity human expert feedback reference call for development fundamental that encompass aspects an objective tractable manner. In paper, we propose one such approach, which novel framework computational pattern analysis ancient data-driven, multilevel, self-sustaining, learning-based, takes advantage large quantities unprocessed available. Unlike many approaches, fast-forward feature vectors, our innovative represents new perspective on task, starts from ground zero problem, definition objects. addition, it leverages data-driven mining relations among objects discover hidden but persistent links between them. The problem addressed at three main levels. At lowest level, images, tackles automatic, enhancement restoration document images using spatial, spectral, sparse, graph-based representations visual second transliteration, directed graphical models, HMMs, Undirected Random Fields, spatial models used extract text manuscript reduces dependency experts. Finally, highest network (from patches words writers) involves search `social networks' linking Considering approach under umbrella Visual Language Processing (VLP), hope will be further enriched by community, form insights contributed various