Visualization

Abstract

The images that best match a query should be displayed to the user in a useful way, that shows immediately if there are any relevant images in the returned set. Traditional image retrieval systems display the returned images as a list, sorted by dissimilarity to the query. At this point, the user can examine the images one at a time, and decide the next action. While this might suffice if the correct images are in the returned list, this is usually not the case, even when the user has as exact image in mind. In many cases, the user has only a vague idea of what he is looking for. In this cases, the display should convey information as to what kind of images were returned by the query, what (if anything) went wrong, how to refine the query in order to get closer to the desired images, and whether or not to terminate the search.

One picture is worth a thousand words.

—Fred R. Barnard

If a picture is worth a thousand words, a picture of an image database is worth a whole book.

—Carlo Tomasi