Comparison of spatial relation definitions in computer vision

Humans are quite adept at recognizing and labeling regions and objects in visual scenes. One of the cues for such labeling is the spatial relationships exhibited among the regions. This is usually coupled with the interpreter's understanding and expectations of scene content. For example, it is normally the case that, in a natural outdoor scene, the sky should be above the trees and that vehicles should be on a road. Context plays a very important role in the interpretation of an image. This determination of spatial relations has been a difficult task to automate. There have been several attempts at defining spatial relationships between regions in a digital image, most recently, with the use of fuzzy set theory. In this paper, we examine three methods for defining spatial relations to gain insight into this complex situation.