Abstract
Introduction: Previously we have shown that human observers use mostly lower order features like the average direction of the light source to determine whether two objects are illuminated in the same way. In this experiment we varied the diffuseness of the light source and the number of light sources while keeping the average direction of the light constant, in order to find out whether these aspects of the light field can be used effectively. Methods: We chose three objects from the ALOI database (Geusebroek et al. , IJCV, 2005), a teapot, an orange and a tennis ball. The objects in the database are photographed with light sources from five different directions. We combine these images with different weights to create our stimuli, either with a single light source that varies in diffuseness in three steps, or with two light sources that vary in their separation in three steps. Observers are presented with an orange, a teapot and a tennis ball on every trial, and have to indicate which of the three objects is illuminated differently from the other two (odd-one-out paradigm). We used all possible combinations of the six light source conditions. Results: Although observers complained that the task was difficult, their performance was nevertheless significantly above chance for most combinations. Moreover, results could also be predicted fairly well by the average pixel differences between two stimuli of the same object, suggesting that observers were able to disregard the differences in the image due to the difference between shape and form of the objects, and focus on the differences caused by the change in illumination. Conclusion: Observers are able to extract information about diffuseness and number of light sources from a scene, in spite of the large differences in shape and material of objects.