Image Pre-Compensation to Facilitate Computer Access for Users with Refractive Errors
Miguel Alonso, Armando Barreto, J. Gualberto Cremades · 2003 · Proceedings of the 6th International ACM SIGACCESS Conference on Computers and Accessibility (Assets '04) · doi:10.1145/1028630.1028653
Summary
This paper presents a novel approach to improving computer access for people with refractive visual errors by pre-compensating displayed images using knowledge of the user's specific optical aberration. The core insight is that the human eye can be modelled as a Linear Time Invariant (LTI) optical system whose blurring effect is characterised by a Point Spread Function (PSF), which can be derived from the eye's wavefront aberration function. If the PSF is known, a mathematical inverse — deconvolution — can be applied to images before they are displayed on screen, so that when the blurred image passes through the aberrated eye, the two transformations cancel out and the retinal image approximates the original undistorted content. Unlike optical corrections (glasses, contacts, LASIK) which are limited to low-order aberrations, this software-based approach can theoretically compensate for high-order aberrations characteristic of conditions like keratoconus. The authors use wavefront analysers (based on the Hartmann-Shack principle, now common in ophthalmology offices) to measure the individual user's wavefront aberration, then apply Wiener filtering (a regularised inverse filter that handles noise and near-zero values in the optical transfer function) to generate pre-compensated images that are displayed on the computer monitor.
Key findings
The approach was evaluated with 14 subjects (28 eyes) using a -6 diopter spherical lens to simulate a severe refractive error, viewed through a digital camera to capture what the eye would perceive. Subjects read simulated Bailey-Lovie logMAR eye charts under three conditions: standard (no lens), blurred (with lens, no compensation), and pre-compensated (with lens, compensated images). The blurred condition severely degraded performance, with mean logMAR acuity dropping from 0.33 (standard) to 1.37 (blurred), corresponding to visual efficiency plummeting from 81.39% to just 1.92%. Pre-compensation substantially improved this to a mean logMAR of 0.67 with 51.69% visual efficiency — a statistically significant improvement (t(27)=64.89, p<0.01). While pre-compensation did not fully restore performance to the unimpaired baseline, it recovered approximately half of the lost visual efficiency. The standard deviation of logMAR measurements also decreased from 0.05 (blurred) to 0.04 (pre-compensated), indicating more consistent performance. A noted limitation was contrast loss in pre-compensated images due to intensity scaling required to fit the modified pixel values within the display's dynamic range.
Relevance
This paper proposes a fundamentally different approach to visual accessibility: rather than modifying the interface (enlarging text, increasing contrast) or correcting the eye (glasses, surgery), it modifies the displayed image to counteract the eye's specific optical aberration. This approach is particularly significant for users with high-order aberrations from conditions like keratoconus, where conventional optical corrections are inadequate and screen magnification alone cannot restore image clarity. As wavefront analysis becomes increasingly routine in eye care (now standard in many optometry practices), the prerequisite measurement of individual aberrations is readily available. For contemporary accessibility practitioners, this work suggests an underexplored dimension of visual accessibility: personalised display compensation based on individual optical characteristics, which could complement existing approaches like magnification and contrast enhancement. The principle of pre-compensating content based on known perceptual characteristics — modifying what is displayed so that what is perceived matches the intended information — could extend beyond optics to other sensory channels, such as audio pre-processing for users with known hearing profiles.
Tags: low vision · visual impairment · refractive error · image processing · visual acuity · computer vision · personalization · display technology · assistive technology