Abstract: | In this project, we investigate attention driven iconic, or picture, matching. In doing so, we review some of the material concerned with iconic vision and discuss various approaches to matching in the iconic domain. We follow this with a description of our own matching algorithm. This consists of two parts: a psychologically motivated attention system that locates objects of interest, and a multi-variate matching algorithm that identifies the located objects. We discuss this system and present results obtained from both artificial and natural scenes. We conclude that although the system is slow, it nevertheless represents a promising, more biologically plausible, approach to vision.
|