GTC Silicon Valley-2019: Sixty Milliseconds To Get It Right: Gaze-Contingent Rendering and Human Perception
Note: This video may require joining the NVIDIA Developer Program or login
GTC Silicon Valley-2019 ID:S9960:Sixty Milliseconds To Get It Right: Gaze-Contingent Rendering and Human Perception
Rachel Albert(NVIDIA)
Truly immersive virtual reality demands a display resolution and field of view that matches the human visual system. Although VR headsets continue to improve, meeting this challenge pushes the limits of per-frame computation time, bandwidth, power, and heat. We'll discuss gaze-contingent rendering, which offers an equally immersive experience at a lower cost by taking advantage of the idiosyncrasies of human peripheral vision. The spatial resolution of human perception decreases rapidly as a function of eccentricity relative to center gaze, allowing a reduced fidelity image to be displayed in the periphery without the user noticing. We'll describe our psychophysical experiment to determine the latency requirements of such a system and discuss our findings showing that state-of-the-art rendering techniques with system latencies up to 50-70 milliseconds were undetectable at the 50 percent threshold level.