Abstract
Peripheral vision exhibits different information processing properties compared to foveal vision (e.g., acuity, crowding, flicker sensitivity, color perception). Gaze-contingent rendering (also called foveated rendering) is designed to exploit the limitations of the peripheral visual system in order to decrease bandwidth, reduce power consumption, and speed up rendering (Guenter et al., 2012). This technique is already being explored in industry (Google & LG’s 1443 PPI display, Fove, and Varjo) as a solution to meet increasing display demands. Previous evaluations of gaze-contingent rendering techniques have focused on detecting artifacts in natural images (Patney et al., 2016; Albert et al., 2017), but it is unknown what kind of effect this type of image degradation may have on observer behavior and performance in other contexts such as reading. Text poses a particularly difficult challenge for gaze-contingent rendering since it contains both high contrast and high spatial frequencies, causing visible flicker and aliasing artifacts when down-sampled in the periphery. Peripheral vision is also known to play an important role in saccade planning while reading (Yang & McConkie, 2001). We measured subject performance under gaze-contingent rendering using a reading task. On each trial, observers were instructed to read a passage and answer whether a particular word appeared in the text. We conducted multiple experiments varying the amount of peripheral degradation across several peripheral rendering techniques, including subsam-pling and Gaussian blur. We recorded reading speed, eye movements, and performance on the task. Results suggest that both the rendering technique and the amount of peripheral degradation influence reading behavior. We demonstrate that performance-based metrics such as reading behavior are an important tool for evaluating peripheral rendering techniques. Furthermore, we advocate for developing new rendering methods to improve the effectiveness of gaze-contingent rendering for text-based imagery.