RECOLOR: A MOBILE VISION ASSISTIVE TOOL FOR COLOR VISION DEFICIENCY USING ADAPTIVE DALTONIZATION AND IMAGE PROCESSING ALGORITHMS

Abstract

Color vision deficiency (CVD) affects roughly 1 in 12 males, yet mobile tools for people with CVD usually address a single function—screening, simulation, or enhancement—and the Philippine National Vision Screening Program does not currently cover CVD. To design, implement, and evaluate ReColor, an Android application that combines a digitized Ishihara screening module, adaptive color enhancement, CIELAB-based color identification, and GPU-accelerated CVD simulation. ReColor was developed through a Modified Agile–Spiral process (15 sprints). We deployed four deterministic pipelines: full-image daltonization by error redistribution, HSV hue rotation as a comparative baseline, CIELAB nearest-neighbor color naming, and a Skia (SkSL) shader for simulation; we trained a U-Net segmentation model as a non-deployed research asset. We evaluated algorithms against synthetic and standard ground truths (verified confusion pairs, X-Rite ColorChecker Classic 24, 30 W3C named colors, and a float64 CPU reference) through system testing and acceptability ratings from 5 participants with CVD, 30 general users, 10 IT experts, and 3 eye-care professionals. Daltonization produced larger discrimination gains than hue rotation for every CVD type (mean +6.91 vs +1.88 ΔE₀₀ across 156 verified confusion pairs), although neither method reached SSIM > 0.90 for most non-neutral test images. The color identifier classified 54 of 54 reference colors correctly, and the GPU shader deviated from the CPU reference by a mean ΔE₀₀ of 0.027. All executed system tests passed. Mean acceptability on a 4-point scale was 3.92 (clinical experts), 3.60 (general users), 3.38 (participants with CVD), and 3.33 (IT experts); participants with CVD rated the enhancement module lowest (2.73). The Philippine Eye Research Institute (PERI) approved the Ishihara module with revisions; its diagnostic accuracy has not yet been established. ReColor is a feasible integrated screening-and-assistance platform. Evidence of functional benefit for people with CVD is preliminary (n = 5) and rests on perception ratings rather than task performance. Next steps include a clinical pilot with normal-vision controls, behavioral testing of enhancement, and on-device latency benchmarking.