Predicting the Perceived Interest of Object in Images

This paper presents the results of a psychophysical experiment and an associated algorithm designed to compute the perceived interest of objects in images. We measured likelihood functions via a psychophysical experiment in which subjects rated the perceived visual interest of each of 408 objects in 100 images. These results were then used to determine the likelihood of interest given various factors such as size, location, contrast, color, and edge-strength. The resulting likelihood functions are used as part of a Bayesian formulation in which perceived interest is inferred based on these factors. Results demonstrate that our algorithm can perform well in predicting perceived interest.