New Metasurface Tech Achieves Accurate 3D Depth Sensing

Passive binocular stereoscopic ranging determines object depth from the geometric disparity between laterally separated views. Unlike active techniques such as time-of-flight or structured illumination, which rely on temporal delays or projected patterns, passive stereoscopy uses triangulation to keep hardware energy-efficient and compact. However, conventional Gaussian imaging systems suffer when axial defocus reduces spatial intensity gradients, causing severe errors in feature localization and depth estimation. A team of researchers has now overcome this limitation by introducing a flat-optics architecture that harnesses subwavelength light control.

The study, published in ACS Photonics, details a polarization-multiplexed binocular metasurface imaging system capable of simultaneous edge-enhanced imaging and passive depth ranging. The platform relies on two identical metalenses separated by a fixed baseline of five millimeters. Each metalens uses a spin-multiplexing strategy built on a specialized meta-atom library featuring a height of 800 nanometers and a period of 350 nanometers, ensuring high transmission efficiency exceeding 90 percent for orthogonal polarization states at a design wavelength of 520 nanometers.

Dual-Mode Polarization Control for Edge Enhancement

The core innovation of the system lies in its ability to switch operation modes solely by altering the incident polarization state of light, avoiding any mechanical adjustments. Under left-handed circularly polarized illumination, the metalens applies a hyperbolic phase profile with a focal length of 20 millimeters, producing a Gaussian focal spot for standard bright-field imaging. When the illumination is switched to right-handed circularly polarized light, the metalens applies a composite phase profile that introduces a spiral phase term with a topological charge of one, generating an annular vortex focal field.

This vortex distribution carries orbital angular momentum, suppressing uniform intensity components while amplifying responses to spatial gradients. This phase-contrast mechanism ensures that structural edge information remains robust even when imaging planes experience defocus. By capturing both bright-field and phase-contrast images from each viewpoint, the system generates complementary datasets. Subtracting the aligned images removes uniform background noise and isolates sharp boundary features.

Stereoscopic Ranging and Quantitative Performance

To test the stereoscopic capabilities, the researchers constructed a dual-channel binocular setup using two laterally separated metalenses. A square pattern served as a distant reference object at 450 millimeters, while a cross-shaped target was moved across discrete axial positions ranging from 15 millimeters to 200 millimeters relative to the reference. As the target shifted closer, defocus and magnification changes blurred the raw bright-field views, but the differential processing successfully preserved clear boundaries.

Stereo matching algorithms compared the extracted edge features from the left and right imaging channels. The lateral displacement between these matched boundaries provided the disparity needed to calculate relative depth. Quantitative analysis revealed a strong linear correlation between the experimentally estimated depth differences and the ground truth values. Across the entire working range, the average relative depth reconstruction error remained at 4.9 percent, with higher accuracy dropping to a mere 0.88 percent at larger depth separations. Furthermore, the standard deviation of the depth estimates stayed consistently low between 1.8 and 2.8 millimeters.

Implications for Compact Machine Vision

Beyond geometric test targets, the team verified the platform’s utility on biological specimens by imaging onion epidermal cells. The dual-mode differential processing successfully highlighted cell walls and local surface features that are often obscured in standard bright-field microscopy. Because the setup achieves near-diffraction-limited focusing in both modes and operates effectively with partially coherent LED sources, it presents a practical route toward biocompatible imaging.

This multifunctional flat-optics approach eliminates bulky optical components while delivering reliable passive depth sensing. The advances described by the research team may soon impact portable three-dimensional sensing, machine vision, autonomous robotics, and advanced biomedical diagnostic tools. Future developments will likely target broadband operation and wafer-scale integration to further expand the commercial viability of metasurface stereoscopic systems.

The research was published in ACS photonics on October 24, 2026.