The theoretical space of visual imagery is effectively infinite. But the images cast on our retinas during daily life have many common features. Statistical regularities in natural visual signals can be exploited by the visual system to efficiently and accurately perceive the world. We study the statistics of natural images and examine their relevance for visual encoding and perception.
Seeing in 3D is an example of a particularly challenging perceptual task. The visual system does not have direct access to 3D information about the environment, so it must be inferred from a variety of visual cues, such as linear perspective and binocular disparity. We study how these cues are used and combined in our perception of the 3D world, and how this perception is shaped by our experiences.
The creation of convincing and practical computer graphics and augmented/virtual reality experiences can be aided by studying the human visual system. We seek to understand when realism is required for accurate perception and comfort, and when it's okay to take shortcuts. We also examine how emerging visual displays can be used to provide assistive technology when vision is impaired.