Tuesday, October 27, 2015

Xerox PARC Develops Hyperspectral Imager

GlobeNewsWire: PARC, a Xerox company, develops "a hyperspectral imaging technology with the potential to be integrated into just about any existing imaging system. The technology demands a minimal cost and size overhead as it relies on a liquid crystal layer about as thick as a human hair."

"The PARC HSI technology endows any existing image sensor with spectral sensitivity without significantly increasing its cost or size. By sandwiching a liquid crystal layer between crossed polarizers and synchronizing the
drive of the liquid crystal with the camera’s image acquisition, the system performs interferometry between two polarizations of light that travel through the liquid crystal. The interferometric data are analyzed to provide the spectral information. Because the complexity of the device is shifted from hardware to software, the sophistication of full spectral processing is within reach anywhere images are normally taken.

PARC has prototyped its HSI technology by integrating a liquid crystal cell inside a commercial monochrome CMOS camera. The prototype offers the following performance:
  • 640 x 480 spatial resolution
  • Up to 80 degree field of view
  • Acquires 30 independent spectral bands in 0.4 seconds
  • Wavelength range 400 nm to 1100 nm
  • F/1.8 max aperture

An open-access PARC paper "Hyperspectral imaging with a liquid crystal polarization interferometer" by Alex Hegyi and Joerg Martini is published in Optics Express, Issue 22, Vol 23.


  1. Time for Apple to take another trip to PARC to "borrow" some ideas. It worked quite well last time.

  2. So after reading the paper it would appear that this is great for this single use but limited by narrow apertures and the need to take 200 frames to compute the fourier transform. Still very useful to have.

    1. Actually, our method is designed to specifically overcome narrow apertures and we can get to f/1.8 or so with our current prototype. We are taking more frames than we need right now, and in general the spectral resolution trades off with the number of frames acquired.

    2. Does the aperture scale with sensor size, ie would it still be 1.8 for larger sensors and does sensor size make a difference?

    3. The choice of aperture is independent of the sensor size. A bigger sensor would allow you to collect more light at a potentially higher spatial resolution (as usual), and sensor size does not affect the spectral properties of the sensor.

    4. Thanks. I was wondering whether larger sensors would be limited by switching speed of the liquid crystal but the paper indicates its the frame rate of the sensor that is the limit so I guess this is not an issue.

  3. What's the width of the photographed band?

    1. The prototype in the paper images from 400 nm to 1100 nm using a silicon CMOS sensor, but the technology is fundamentally compatible with image sensors all the way out to the long-wave IR.

    2. I mean what’s the band width of a set wavelength - like for the 480nm photo what's the cut on/off wavelength.

  4. Please see Fig. 6 of our paper:

  5. 31.5 nm FWHM is a bit to much but if it will allow ~15nm or less FWHM then it can be a technology for astronomical narrowband imaging.

  6. It would be great to have a (free) Prototype (with USB Type-C Interface) after it is a bit more developed.

    Some interesting additional features it might include could be:

    1.) 'Tricorder' Mode which captures the spectrum in narrow bands permitting external processing to perform an analysis and guess what you are pointing it at.

    2.) High speed Shutter capable of snapshoting Laser startup or capturing extremely brief samples of very intense sources.

    3.) Wafer Level Optical component Lens offering: (3A) GRIN with slightly variable focal length, (3B) micro-Fresnel lenses, (3C) agile beam steering using binary optical microlens arrays.

    4.) RGB Mode allowing capture of both a conventional RGB image (like a Camera) and capable of shifting the center frequency to capture any three Bands (so they could be viewed as if humans had a different portion of the Spectrum seen as visible light).

    5.) Etc.

    I hope that was explained both simply and technically enough to appeal to most of our audience.

    Some or all of those features would be useful to Xerox in existing products.


All comments are moderated to avoid spam.