A Polarization Contrast Retina That Uses Patterned Iodine-Doped PVA Film

We present the design of a 1D polarization contrast retina. The retina employs two parallel, linear arrays of 29 photodiodes as sensing elements. Polarizing material is placed on the focal plane so that each array of photodiodes receives linearly polarized light whose e-vector components are orthogonal. On chip analog subthreshold MOS circuit computes the polarization contrast. We also discuss how we can process iodinedoped poly(vinyl alcohol) linear polarizers using standard microfabrication lithographic techniques to delineate features with resolution comparable to pixel sizes of standard imagers. 1. Polarization Vision The retinae of most insect and certain vertebrate species are sensitive to polarization in their environment, but humans are blind to this property of light. Currently, scientists use polarization information to perform difficult computer vision tasks, such as image segmentation, object orientation, material classification, and atmospheric and solar research. Biologists use polarimeters to investigate behaviors of animals – vis-à-vis polarization – in aquatic habitats. Complex vision tasks are made tractable with polarization vision techniques because polarization is a more general descriptor of light and contains physical information about reflecting objects in a scene that traditional intensity based sensors ignore. The electric field (e-vector) of light can be expressed as the superposition of two orthogonal components, Ex and Ey, and can be written as E → = ˆ xAe − j ( t −kz+ a ) + ˆ y Be j ( t −kz + b ) , (1) where is angular frequency, k is the wavevector, and is phase. Polarized light results when the phase between orthogonal components is deterministic. In the case of 0° phase difference, light is linearly polarized; in the case of non-zero phase difference, light is elliptical polarized. While polarized light sources are rare in nature, almost all reflected light is partially linearly polarized, which means that reflected light consists of an unpolarized Fig 1 Photomicrograph of the 1D PCR taken through a polarizer oriented at 90°. Photodiodes under the 90°orientation film are visible; photodiodes under the 0°orientation film are blocked by the canceling effect of two orthogonal polarizers. component and a polarized component that is linearly polarized. Intensity based sensors lump the orthogonal evector components into one intensity term by summing the energies of Ex and Ey. Polarization sensors, on the other hand, use these components to extract important physical features such as specularities, occluding contours, and material properties. We describe a 1D polarization retina that can be used as a polarimetric sensor for real-time, automated vision tasks. Our polarization contrast retina (PCR) uses linear polarizers on the focal plane and on-chip analog, subthreshold circuitry to compute polarization contrast. 2. 1D Polarization Contrast Retina Polarization contrast measures relative polarization phase and partial polarization in a scene. We define polarization contrast using the following equation: polarization contrast = TR90 − TR0 TR90 + TR0 , (2) where TR0 and TR90 represent the transmitted radiance through orthogonal polarizers. The 1D PCR uses linearly polarizing film that consists of two orthogonally polarizing regions placed directly over two parallel arrays of photodiodes (Figure 1). Linear polarizers cannot distinguish between linear and elliptical polarization; therefore, the PCR cannot distinguish between elliptically and linearly polarized light. The arrays are addressed such that the current outputs of two photodiodes that lie across from each other under different polarizers are routed to an analog circuit that computes polarization contrast (Figure 2). The circuit operates on the two currents, which are proportional to the transmitted radiance through the 0°-orientation and 90°-orientation polarizers, and produces one output current that represents the polarization contrast, IPC. An analog, current-mode circuit computes polarization contrast (Figure 3). The circuit is implemented in subthreshold MOS. Ideally, when MOS transistors are used in the subthreshold region, drain current is exponential with Vgs. Using this property of transistors in subthreshold and the translinear principle, we can write Vgs1 + Vgs3 = Vgs 2 +Vgs4 ⇒ I1I3 = I2I4 . (3) Since the MOS transistors share common substrate and source terminals, the translinear behavior of this circuit is exact. Kirchoff's current law gives us the following additional equations: