The preservation, analysis, and management of collections of valuable historical and cultural materials via digital imaging techniques is an emerging field significant to the society at large. The benefits of using visual and multimedia analysis techniques in this domain are manifold. Automatic content-based analysis methods can provide scholars in the humanities and the arts with tools to facilitate some of their more rutinary work - consulting, annotating, cataloging - but also to obtain new insights about specific theories through the recognition and discovery of patterns within and across pieces in a collection. Furthermore, these techniques can be used to create systems to access and retrieve digital versions of the materials, with the general public and education as the intended audience and goal, respectively. Finally, the automatic analysis is challenging, considering the nature of the media: complex content, limited amount of data, and degraded quality due to the effect of time. This domain thus constitutes an excellent testbed to develop robust analysis methods that can have general applicability to other media sources.
The project addresses the problem of analyzing digital imagery of cultural heritage materials through the development of new and principled analysis methods of wider applicability, integrating computer vision, information retrieval, and machine learning techniques. The project aims at advancing the state-of-the-art on representation, categorization, and matching of visual entities (for example, objects), and on discovery of salient visual patterns (visual data mining) in complex image collections, relying on robust invariant visual features and probabilistic models.
While the developed visual analysis methods aim at being generic, they will be showcased in a tool to support specialists' needs for a unique and challenging cultural heritage data source: the collection of ancient Mesoamerican pictorial documents from the National Anthropology and History Institute (INAH) in Mexico, which constitutes an invaluable source of information about the history, language, and beliefs of the indigenous cultures that inhabited Mexico before the Conquest of America. A fundamental research issue is the exploitation of prior domain knowledge and metadata, which will be investigated through a close collaboration with scholars at INAH. As a result, the project will join scientific research in the humanities, through the participation of archaeologists as a source of expertise and as users of the technology, and in computing, by targeting a challenging case for developing new methods for analysis of collections of visual information.