Predicting the Perceived Interest Of Objects in Images

Abstract

This thesis presents an algorithm designed to compute the perceived interest of objects in images based on results of a psychophysical experiment. We measured likelihood functions via a psychophysical experiment in which subjects rated the perceived visual interest of over 1100 objects in 300 images. These results were then used to determine the likelihood of perceived interest given various factors such as location, contrast, color, luminance, edge-strength and blur. These likelihood functions are used as part of a Bayesian formulation in which perceived interest is inferred based on the factors. A block-based approach is also proposed which doesn't need segmentation and is fast-enough to be used in real-time applications. Our results demonstrate that our algorithm can perform well in predicting perceived interest.