Determination of the Fundamental Image Categories for Typical Consumer Imagery

Many tasks in imaging science are image-dependent. While a particular dependency might simply be a function of certain physical attributes of an image, often it is closely related to the perceived semantic category. Therefore, a thorough understanding of image semantics would be of substantial practical value. The primary goal of this research was to determine the fundamental semantic categories for typical consumer imagery. Two psychophysical experiments were performed. Experiment I was a Free Sorting Experiment where observers were asked to sort 32 1 images into piles of similar images. Experiment II was a Distributed Experiment conducted over the internet which used the method of triads to collect similarity and dissimilarity data from 321 images. Due to the large number of images included in the experiment, the method of non-repeating random paths was employed to reduce the number of required responses. Both experiments were analyzed using multidimensional scaling and hierarchical cluster analysis. The Free Sorting Experiment was also analyzed using dual scaling. The results from all three methods were compiled and a set of 34 categories that proved to be stable across multiple methods of analysis was formed. A multidimensional perceptual image semantic space has been suggested and advantages to utilizing such a structure have been outlined. The 34 fundamental categories were represented by 10 perceptual dimensions that described the underlying perceptions leading to categorical assignments. The 10 perceptual dimensions were humanness, artificialness, perceived proximity, candidness, wetness, architecture, terrain, activeness, lightness, and relative age