What Are Eyelashes?

Eyelashes are the rows of sensory hairs that line the margins of the upper and lower eyelids, serving both a protective biological function and a significant role as a source of noise and occlusion in optical biometric imaging systems. In engineering contexts, eyelashes appear most prominently in iris recognition and periocular biometrics, where their presence over the iris region degrades segmentation accuracy and must be detected and masked before feature extraction proceeds.

The biological structure of eyelashes includes the lash follicle, the lash shaft, and the sebaceous glands that lubricate the shaft. Each eyelid carries between 75 and 200 lashes, arranged in two or three rows, with upper lashes typically longer and more numerous than lower lashes. In clinical and biomedical engineering contexts, abnormalities in lash growth direction, such as trichiasis, can cause corneal abrasion and are detected using slit-lamp imaging and automated segmentation tools derived from the same computer vision techniques used in biometric systems.

Eyelash Detection in Biometric Systems

The primary engineering challenge associated with eyelashes is their occlusion of the iris. In iris recognition systems, the iris must be segmented from the surrounding eye region before its unique texture pattern can be encoded and matched. Eyelashes crossing the iris boundary introduce dark, curved artifacts that corrupt the iris code if left unmasked. A novel eyelash detection approach published in PubMed categorizes lashes into two types: separable lashes, identified by local intensity minima along radial scan lines, and multiple overlapping lashes, classified by measuring pixel-region mean and standard deviation against learned thresholds.

Early segmentation pipelines used the Hough transform to locate the iris and pupil boundaries and then applied connected-component analysis to label lash-contaminated pixels. More recent approaches use convolutional neural networks with dual-decoder architectures: one decoder estimates the geometric boundary of the iris and a second decoder specifically predicts noise-element masks that include lashes, stray hairs, and specular reflections. Both outputs are combined before the iris texture is sampled, reducing false-reject rates in systems deployed under uncontrolled lighting.

Eyelash Segmentation in Periocular Imaging

Beyond iris recognition, eyelashes are a useful feature in periocular biometrics, where the entire region surrounding the eye serves as a recognition template. Periocular identification is valuable when the full iris is not visible, such as in images captured at a distance or when subjects wear glasses. In this context, lash geometry, including lash count, curvature, and spatial distribution, provides additional discriminative information that complements iris texture and skin texture features.

Accurate lash segmentation also appears in ophthalmic imaging workflows. Anterior-segment photography and slit-lamp video capture must exclude lash shadow artifacts before clinicians measure limbal boundaries or track corneal surface changes over time. Research into iris recognition as an emerging biometric technology has highlighted lash occlusion as one of the primary sources of segmentation error, prompting sustained work on robust detection algorithms. Automated lash masking pipelines reduce manual annotation workload in large ophthalmology datasets.

Applications

Eyelashes, and engineering methods developed to handle them, have applications in a range of fields, including:

• Iris and periocular biometric identification systems
• Ophthalmic imaging and corneal diagnostics
• Eye-tracking systems that require clean segmentation of the pupil and iris
• Driver monitoring and fatigue detection, where lash occlusion affects gaze estimation accuracy
• Forensic image analysis and identity verification from surveillance video