Foveated rendering is an optimization technique for virtual reality (VR) headsets that utilizes eye tracking and the fact that peripheral vision senses less detail than the center of vision, the fovea centralis of the eye. Overall improved graphical performance can be achieved by reducing the rendering workload with decreased image quality and complexity outside of the observed center of the stereo images, and is essentially undetectable by a viewer.
The an example that combines foveated imaging, eye tracking, and augmented reality headsets was presented in a paper, Perceptually-Based Foveated Virtual Reality, that was published by NVIDIA Research in 2016. One measured performance improvement was the reduction of pixel shading performance by 2-3x. The FOVE startup sought to include foveated rendering in their headset and described benefits such as increased center detail, higher frame rates and rendering speed.
As a happy owner of the HTC Vive VR headset I can still identify several areas for improvement that might be helped via foveated rendering such as low perceived detail due to display resolution, refresh rate, and graphics quality level that is the result of limitations in current display technology and pixel density, available data throughput from GPU to headset via DisplayPort and HDMI standards, and available processing power on the computer itself. I’m looking forward to the next generation of VR headsets when foveated rendering can be made available in conjunction with those other processing and display improvements.
References: