To investigate the human ability to discriminate between simple volumetric objects differing in either elemental shape or relational structure a prototypical componential model representation for solid objects was initially developed. There were three principal subsections a set of global properties, a set of shape descriptions of the individual components, and a description of pairwise relations between components. Measures of similarity between objects were calcuated on the basis of an algorithm derived from relational database metrics. A picture/picture, same/different paradigm was employed to determine the ability of human subjects to discriminate between objects with single variations in either the shape or the relative position of components. Significant differences in correct response times and error rates were observed depending on the type of shape variation of individual components, and on the degree of lateral shift between them. Furthermore, a spatial order of processing was observed which could not be accounted for by individual component characteristics. This showed a primary effect from top to bottom of the presented stimulus. The findings are discussed in terms of models of object recognition and of other work on spatially ordered processing. © 1993.
|Number of pages||17|
|Journal||Image and Vision Computing|
|Publication status||Published - Jul 1993|
- componential models
- human vision
- object discrimination