Querying and retrieving pictorial data using semantics induced by colour quality and arrangement

In this paper a novel approach is featured to querying painting databases on the basis of the semantics implied by the use artists make of colours. A symbolic query language, called "colour description language" has been designed to express high-level chromatic features of a painting, which cannot be obtained through a set of concrete examples. A visual extension of the language is then presented which introduces a feedback between the sentence composed and its pictorial meaning. The language has been used as the high-level query paradigm of a prototype automatic painting database.