Informing Content- and Concept-Based Image Indexing and Retrieval through a Study of Image Description
This research reports on a Web survey of visual resource experts. The research uses the data supplied by the experts to discuss if content-and concept-based image indexing and retrieval approaches may be improved by examining features relevant to both. Image descriptions were gathered from image professionals via a Web survey. The descriptions were analyzed to test whether color images evoke a denser textual description than grayscale images, and whether there is any significant variation in the words used to describe them. Analyses of data resulting from the survey of image description indicate that image professionals describe images with high-level concepts rather than low-level features, and that color does not affect the number or type of words used in their descriptions. This work may prompt future research with other building blocks of the systems in order to better integrate the research done by computer and library scientists in this area.
School:University of North Carolina at Chapel Hill
School Location:USA - North Carolina
Source Type:Master's Thesis
Keywords:pictures databases indexing information retrieval subject headings art
Date of Publication:02/17/2005