Retinotopic Foveated Rendering

Figure 1: We propose retinotopic foveated rendering (RFR) based on the horizontal-vertical asymmetry (HVA) and vertical-meridian asymmetry (VMA) that are evident from advancements in retinotopy research. We first derived the asymmetric model for cortical magnification factor (CMF), which is proportional to the visual acuity. The comparison of the radially symmetric CMF model in typical FR and the radially asymmetric CMF model in the proposed RFR is illustrated in the left column, indicating the overestimated visual acuity near the vertical meridian in existing FR methods. The asymmetric CMF model is derived from the fMRI data of the early cortex, which defines the asymmetric rendering precision along the horizontal, lower-vertical, and upper-vertical meridian at ratios of 0.72:0.57:0.32. The across-meridian asymmetry is achieved through interpolation in terms of the quasi-linear degradation nature of the visual acuity from the horizontal meridian to the vertical meridian. As a result, RFR keeps consistent rendering precision in the horizontal meridian while eliminating excess rendering in the upper and lower FOV, as shown in the zoom-in images enclosed in red, orange, and blue boxes in the middle column. Notice that the degradation of image quality in RFR is unperceivable since it follows the radially asymmetric regression of CMF. We validated the proposed RFR method by integrating the asymmetric computation model with both polar and rectangular mapping-based foveated rendering. The results show that RFR methods reduce shading in the fragmented shader by 27.2% on average compared with typical FR methods, resulting in the increase in frame rate by 17.45%.