Image-Based Identification of Ophthalmic Lens Optical Characteristics: A Benchmark of Local Texture Descriptors for Non-Destructive Optical Inspection

Description

Abdelilah Errachidi and colleagues ask whether a photograph can sort spectacle lenses by their invisible specs, in this benchmark study from twenty twenty-six. A lens bends and tints the view through it in ways too subtle for the eye, leaving a micro-texture signature against a controlled target. Reading that signature would allow inspection that never touches or damages the lens.

The team built two image sets and ran a field of classic texture descriptors, the pattern-counting family led by local binary patterns, under one shared protocol. Several modern pattern variants recognized lens categories, by refractive index and coating type, with excellent accuracy. Deep neural features scored highest overall, but the handcrafted descriptors came close with no training at all. The authors offer the benchmark, the first for this task, as a guide for inspection systems that must be cheap, stable, and light on data.

Catalog record

Source: https://europepmc.org/articles/PMC13611072