Compared to Red Green Blue (RGB) patterned colour imaging sensors, polarization imaging sensors deliver another set of information about objects reflecting light. As the researchers explain in the introduction of a paper “Bioinspired polarization imager with high dynamic range” published in the Optica journal, the polarization state of light can act like a memory foam by “remembering” the intrinsic properties of the media or objects that light has encountered in previous optical interactions, this includes information about their three-dimensional shape, surface roughness, and material or tissue structural composition. The authors think that a cheap high dynamic range polarization imager would have its place on-board cars to provide critical information during hazy or rainy conditions when traditional sensor may fail. A high-dynamic-range is also desirable to capture scenes where the illumination can easily vary by several orders of magnitude, when exiting a garage or a tunnel or when cross the light beam from an oncoming vehicle.
We humans don’t have the ability to distinguish the polarization of light but many species do and use this ability for their navigation or for visual communication. The mantis shrimp is often cited as sporting one of the most complex visual systems found in nature, capable of detecting 16 spectral channels and 4 linear and 2 circular polarization channels. What’s more, its individual photoreceptors have logarithmic responses to incident light intensity, enabling a high dynamic range.
A closer look at the mantis shrimp’s compound eyes yields shows closely packed protrusions (microvilli) in the light-sensitive receptors, rotated 45° across two different areas of the eye which makes the peripheral hemispheres of the eye sensitive to two orthogonal orientations of linearly polarized light. This allows the mantis shrimp to detect four e-vector orientations of polarized light.
In order to replicate such a structure, the researchers first fabricated a 384×288 pixel sensor array (with a pixel pitch of 30μm) in a CMOS 180nm process. They then monolithically integrated polarization filters at the pixel level as a 2×2 pattern of pixelated polarization filters offset by 45°, repeated across the whole imaging array. That means four adjacent pixels get four different polarization filters at 0º, 45º, 90º and 135º. In this array, the individual polarization filters were made of stacks of 250nm-tall and 75nm-wide aluminium nanowires, with a 50% duty cycle (laid out to alternatively block and let the light through).
But the polarization filters were not enough. In order to achieve a high dynamic range, the researchers designed the pixel circuitry to operate the photodiode in the forward bias mode, unlike traditional active pixel sensors, which operate it in reverse bias mode. Thanks to a custom 3-transistor readout architecture, the photodiode voltage in the sensor is no longer linearly proportional to the photocurrent or photon flux. Instead it follows a logarithmic response. In effect, the photocurrent measurement gets compressed in the voltage domain as the photon flux increases, making the logarithmic pixel more sensitive to larger photon fluxes than a traditional linear active pixel.
Measurements performed with the monolithically integrated bioinspired imager revealed a dynamic range of 140 dB with a signal-to-noise ratio of 61 dB. These figures are about 600 times greater and 5 times greater, respectively, than the highest figures reported in the literature, according to the authors. What’s more, this instant multiple polarization acquisition device allows snapshot polarization imaging at 30 frames per second.
The researchers have filled for patents on this camera design and have also formed a start-up company, Mantis Vision Inc. to commercialize this new polarization imaging sensor, with automotive applications in mind.
“The next step would be to combine both colour-RGB together with high dynamic range for automotive applications. Right now we are using both sensors for autonomous cars and are evaluating if we need to combine both of them together into a single sensor, which is not difficult to do”, commented corresponding author Viktor Gruev, associate professor of electrical and computer engineering at the University of Illinois.
Indeed, Gruev also shared a paper published last year titled “Bio-inspired color-polarization imager for real-time in situ imaging”, describing the integration of vertically stacked RGB photodetectors with polarizing filters on top, allowing the co-registration of colour and polarization information in one snapshot.
University of Illinois – www.illinois.edu