A Conceptual Framework and Empirical Research for Classifying Visual Descriptors
Corinne Jorgensen, Alejandro Jaimes, Ana B. Benitez and Shih-Fu Chang
This paper presents exploratory research evaluating a conceptual structure, a ten-level Pyramid, for the description of visual content of images. The Pyramid was developed from empirical research in several fields, including Cognitive Psychology and Information Studies. Four research questions are posed and several experiments test the ability of the Pyramid to accommodate spontaneous image descriptions, to accommodate structured image descriptions, to generate structured image descriptions, and to accommodate varying levels of description. Input to the testing includes image descriptions generated by nave users and indexers, images from a web sample and news images, and the Pyramid structure and an Indexing Template. Testing involved mapping and generating a number of image descriptions. A variety of techniques were used to evaluate the conceptual structure and these techniques produced consistent and positive results across a variety of image sets. The authors feel that the Pyramid is a robust conceptualization of visual image content and encourage further work developing this structure for image representation and retrieval.