The Software Lens: Eliminating Prescription Inserts in Near-Eye Displays via Inverse Ray Backtracing and Per-Pixel Directional Emission

Abstract

Users with refractive errors -- roughly one third of the world's population today, projected to approach one half by 2050 -- cannot use consumer VR headsets comfortably without corrective optics. The industry's answer is the prescription insert (RX insert): a custom-ground lens pair that clips between the headset's viewing optics and the user's eyes. Inserts are expensive, slow to procure, single-user, incompatible with shared and enterprise devices, and they consume scarce eye-relief distance inside an optical stack that headset designers are fighting to shrink. We propose eliminating the physical insert entirely and replacing its optical function with software-defined light. Our key observation, developed and validated in a multi-stage ray-tracing simulation of the full optical chain, is that the insert's entire contribution is a deterministic, per-ray redirection of the light bundle that enters the pupil. We capture this contribution by a two-pass procedure: a forward reference pass traces rays from each display pixel through the viewing optic and the prescription lens into the eye; an inverse backtrace pass propagates that corrected ray set backward toward the display with the prescription lens virtually removed, yielding the precise emission direction each display pixel would need in order to reproduce the corrected bundle using the headset's remaining optics alone. We formalize this equivalence in paraxial ray-transfer (ABCD) analysis, describe two display realizations of the resulting primitive -- the dixel, a directional pixel -- a ray-steered architecture (laser + MEMS micromirror, Maxwellian view) and a coherent phase-array architecture, and validate the approach across a three-stage simulation pipeline (interactive 2D ray optics, an exact-Snell MATLAB tracer with a multi-surface schematic eye, and physically based 3D ray tracing), cross-checked to approximately 3 micron RMS agreement in retinal intercepts. We report simulated acuity improvements from approximately 20/250 (uncorrected, -3.00 D myope) to 20/30-20/70 (ray-steered) and 20/20 (phase-array), and extend the architecture to a dual-directional smartphone display for glasses-free, per-eye prescription correction on shared flat panels. This is a simulation-validated architecture paper; we do not claim a hardware prototype.