Inspired by the design of the retina in the human eye, scientists in Scotland have managed to circumvent the usual trade-off between speed and resolution in a single-pixel camera system (D. B. Phillips et al., Sci. Adv. 3, e1601782; 2017).

Single-pixel imaging works by applying a consecutive series of different binary patterns to the optical path of a scene to be imaged and measuring the correlation with data that is synchronously recorded on a single photodetector. However, there is an intrinsic trade-off between the resolution and frame rate of such systems as the number of measurements is equal to the number of reconstructed pixels in the final image.

Miles Padgett, David Phillips and co-workers at the University of Glasgow have now demonstrated a way to get around this limitation by employing binary patterns that, instead of having a uniform resolution, have a spatially varying resolution that resembles the retina of the eye. In particular, the pattern mask (left panel) has a fovea-like region of high resolution that is surrounded by a lower resolution in the periphery. The benefit of the approach is that it allows higher-resolution imaging in a region of interest (right panel) without slowing the frame rate of the imaging to the level that would be needed for a high resolution over the entire field — Padgett's team report a local frame rate enhancement of a factor of 4. Furthermore, the location of the 'fovea' region can be changed at will, allowing the opportunity for object tracking or multiple fovea-like high-resolution regions to be employed.

Credit: HTTPS://CREATIVECOMMONS.ORG/LICENSES/BY/4.0/

“We were working together on a system which used patterns of different scale to allow us to switch between different resolutions. Dave had the idea of doing high resolution in the middle and low around the outside — just like the eye,” commented Padgett. “We think it is of benefit to all single-pixel cameras or similar systems — for example, we have converted our recent time-of-flight system to use the same approach.”

The Glasgow system makes use a white LED torch as a source of illumination and a computer-controlled digital micromirror device with 32 × 32 programmable pixels to create a dynamic binary mask that is applied to the image of the scene prior to detection by an avalanche photodiode (APD). The team says in addition to operation at visible wavelengths, the approach is equally applicable to other wavelength regimes. With this in mind, they have also demonstrated a short-wave infrared (SWIR) single-pixel imaging system that operates in the 800 to 1,800 nm wavelength region by replacing the APD with a SWIR-sensitive detector and using a heat lamp as an illumination source.

Potential applications that could benefit from the approach include machine vision and gas sensing, for example.