Context-sensitive queries for image retrieval in digital libraries
A common notion of vision, consistent with this excerpt from the first paragraph of David Marr’s seminal book , is gained by considering the predicament of a person with a searchlight placed in a pitch-dark room full of unfamiliar furniture. One would hope that, by swinging the beam around, the observer will be able to recognize the objects present in the room (a cat here, an aquarium there, etc.) – a task that no longer appears as daunting as it used to because its computational nature is now better understood [2,3]. There is, however, more to highlevel vision than recognizing and mentally labeling one object after another, just as there is more to our visual world than a list of objects in the field of view that can be ticked off. Unless viewed in the unusual conditions of darkness with the aid of a searchlight, objects present themselves to us embedded in scenes, combined and recombined in a highly variable, yet structured, manner.