| dc.description.abstract | One of the most important processes in the human visual system involves detecting and
understanding edges. Edges allow humans to break a visual scene up into meaningful
chunks of information. Without edges, a visual scene is meaningless. As important as
edges are to human visual perception, how they are detected and classified is not well
understood. This study provides evidence that humans are able to classify edges into
appropriate categories when enough visual information is presented but objects in the
scene are not detectable. In addition, this study shows that regions of interest (RO Is) of a
particular edge type can be clustered according to similarities in structure using a simple
algorithm. This study examines the relationship between image features (i.e. closure,
texture & repetition) and the type or cause of an edge (i.e. albedo, depth, shadow &
specular) in natural visual scenes. Two groups of human subjects were used to carry out
the current study; the cause estimators (CEs) and the feature experts (FEs). The CEs were
asked to state the cause of an edge presented in a ROI. The FEs were asked to label
specific features for the same set of RO Is as the CEs. The first analysis describes the
relationship between image features and the actual cause of the edge in the ROis presented.
The second analysis describes the relationship between image features and the cause
estimation provided by the CEs. This study provides evidence that closure, texture and
repetition may help to inform human observers as to the cause of an edge when limited but
sufficient visual information is available. | en_US |
