Imaging under polarized light provides more surface details of the photographed object compared to capturing it under natural light, as we can trace the different ways, varying degrees of polarized light interact with complex surface details of the object. In previous works, linear and circular polarization filters used in [1]on the camera lens as well as polarization-sensitive image sensors used in [2] have been used to acquire this information. Refractive optics that have traditionally been used for material acquisition using polarized light are bulky and as a result, it is non-trivial to rapidly capture polarized images using polarization filters from the same view angle without specialized sensors. In this work, we leverage nanoscale imaging by designing and implementing a metalens to record two incident polarizations under the same viewpoint and show that we can use polarimetry to solve for the surface details of the captured object. We achieve this by, first, simulating a polarization-sensitive bifocal metalens enabling us to record and verify the far-field profiles of cross and parallel polarized incident illumination under the same viewpoint so that they are resolved separately, as two focus peaks, on the same focal plane but with a small distance between them. We then record the designed metalens’s response to an object lit by this polarized incident illumination to infer the diffuse and specular properties of the surface, thereby allowing us to capture material properties with a nanoscale system as well as deriving more information from the designed imaging system in a single shot.

Photorealistic digital representation of real-world objects in a 3D scene requires the appearance of the object to be faithfully represented virtually. The interaction of the lighting in the scene with the object, parametrized by its material properties and geometry dictates how the object appears to the observer. In order to render the object in 3D, the same object-lighting interactions need to be modeled, thereby requiring the object’s physically based material properties and 3D shape. While the 3D shape is acquired by photogrammetry approaches, such as in [3–5], the object’s material can be acquired by measuring its surface reflectance properties as in [1,6,7]. The rendering algorithm is highly streamlined when the diffuse and specular reflectance of the object are separated as shown in [8]. Several approaches have been studied and developed for diffuse-specular reflectance separation, mainly image intensity-based methods such as[9,10] and polarization separation-based methods such as [11–13]. Authors of [1] used a linearly polarized lighting control rig along with multiview cameras tuned to linearly polarized filters to provide the ground truth polarization-difference imaging data needed to acquire surface geometry and reflectance properties of the object being imaged. In [14], the authors used a single camera [2] with a polarization grid for four polarization angles on the camera image sensor to solve for the properties of the object material. However, these approaches need active lighting and bulky optics-based cameras which are unsuitable for AR and VR applications as the application requires small form factors.

To prove that our designed metalens is in fact able to capture the diffuse-specular separation, we use it to image objects - a non-scattering periodic surface and a scattering irregular surface with known diffuse and specular reflectance. Wile using the field profile of these objects as source illumination we run a FDTD simulation of the metalens to observe the resulting far-field profiles. Both simulations have an incident illumination with S-polarization. We see from Fig. 4 (a) that the scattering object does in fact scatter the polarized light thereby producing two focus peaks on the far-field profile. Since the object was lit with S-polarization illumination, we see a denser energy peak formed on the right side of the x-y plane cross-section.

In this paper, we design a bifocal polarization-sensitive metalens and show that our design focuses the x- and ypolarizations of incident illumination onto separated focus peaks measured by the far-field profile of the metalens. We also show by using two objects, a non-scattering periodic surface and a scattering irregular surface with known reflectance and transmittance, the smooth reflective object the metalens focused the peak at the respective incident illumination polarization, and for the irregular object there was in fact light scattered and it was focused on the opposite peak of the incident illumination. Simulating the entire system with an image sensor rather than a field profile is the next natural progression. We have yet to realize a physical metalens system by fabricating the metalens with a larger diameter. This system can then be further extended to a light array system that is sensitive to more than two polarizations allowing us to achieve material properties using Stokes-Muller analysis with passive illumination.