Abstract
When comparing illumination conditions, human observers mostly extract the direction of the light source from low-level image cues. The question we ask here is how well they are able to distinguish other low-level aspects of illumination, like the diffuseness of the light and the number of light sources. We also investigate what kind of stimulus information is most important for this task. To address this question, we used pictures of a teapot, an orange and a tennis ball from the ALOI database (Geusebroek et al., IJCV 2005) to create 6 illumination conditions for each object. The objects are illuminated from a single direction and varying in diffuseness or from two directions that with varying separation. Observers are presented with all three objects on every trial, and have to indicate which one is illuminated differently from the other two. We measured performance and reaction times on every trial. We also recorded eye-movements to determine what part of the stimulus our participants were looking at to complete this task. Results show that participants performed above chance for most conditions, and there are systematic variations in performance for different conditions. These differences in performance were predicted well by a model that uses differences in image structure in same-object comparisons. This model suggests that participants mostly rely on the information in cast shadows to perform the task. Interestingly, participants primarily look at the shadows (roughly 60% of the fixations), in favor of shading (30%) and highlights (10%). When we look at the reaction times, we see that the pattern of results is similar as that of performance, suggesting that higher performance correlates with shorter reaction times. Moreover, when there is a larger difference in highlight structure the reaction times are shorter.
Meeting abstract presented at VSS 2014