https://cdn.specpick.com/images/photonics/mit_lenseless_image_0.jpg712370

Compressed sensing is an exciting new computational technique for extracting large amounts of information from a signal. In one high-profile demonstration, for instance, researchers at Rice University built a camera that could produce 2-D images using only a single light sensor rather than the millions of light sensors found in a commodity camera.

But using compressed sensing for image acquisition is inefficient: That “single-pixel camera” needed thousands of exposures to produce a reasonably clear image. Researchers from the MIT Media Lab developed a new technique that makes image acquisition using compressed sensing 50 times as efficient. In the case of the single-pixel camera, it could get the number of exposures down from thousands to dozens.

Methods to obtain high-quality images without the use of a high-magnification objective lens, to map light from an object onto a suitable sensor plane, have been making steady progress for some time now. One particular reason for this being, the advances in modern signal processing techniques and the opportunity they provide to shift the emphasis in an imaging operation away from the optical hardware and onto the computational aspects instead.

A project from the Camera Culture Group at MIT Media Lab has now developed a method for lensless imaging that leverages both compressive sensing (CS) - one of the foundational numerical methods of computational imaging - and current breakthroughs in time-resolved optical sensing, a technology which is already key to several group research projects. According to the Group's results, the efficient lensless imaging is possible with ultrafast measurement and compressive sensing, and point towards ways that novel imaging architectures could be put to use in situations where imaging with a lens is impossible.

Apparently, this is the first combination of time-resolved sensing with a single-pixel camera used for detecting reflectivity - effectively, a photography application. Single-pixel systems have been known and investigated for some years, while time-resolved sensing has been used for measuring reflectivity without compressive sensing, as well as in LIDAR to recover scene geometry. But the MIT group has claimed to have developed the missing piece of the puzzle, combining these approaches to recover the reflectivity and albedo of a scene.

As reported in a paper for IEEE Transactions on Computational Imaging, the new approach ultimately made image acquisition using compressive sensing more efficient by a factor of 50. In lensless single-pixel camera systems - which rely on multiple measurements by the same sensor pixel under different illumination patterns, each controlled by a spatial light modulator so as to encode different information into each measurement, the findings could help to reduce the number of exposures typically involved from thousands down to dozens.

The MIT project investigated three distinct, but closely connected, fundamental issues in lensless imaging, starting with how to guide designers towards the best system architectures for lensless applications in general. A new design framework created by the Group provides a set of guidelines and decision tools to suggest how the available resources in a given scenario can be deployed to recover the best image using compressive sensing, and define when the compressive sensing approach is likely to be of most benefit. This need not necessarily involve single-pixel sensing but the framework is intended to answer relevant design questions about the best positions for particular sensors, or lay out the conditions under which an improved time-resolution might be more beneficial than additional detectors.

A second project thread was followed to examine the role of time-resolved signals, and clarify how an improved temporal resolution can reduce the number of individual modulated signals needed to build up a high quality image. The third area of investigation related to the optimization of those individual patterns of modulated light, and ways to squeeze more information out of each one.

Compressive sensing allows to modulate the light in a smarter way. Without it, a large number of measurements to obtain a high-resolution image has to be done, making it time consuming. The geometry of the new system will significantly affect the time-resolved measurements with points closer to the detector being measured first while points further being measured later. Modulating the light allows to accommodate this effect and obtain more information per measurement and thus potentially needing fewer modulation patterns to obtain a full result.

Apart from MIT, a recent indication of the potential value of lensless single-pixel camera systems has come from M Squared and the University of Glasgow, and the development of an instrument to directly image methane gas leaking from a ruptured pipeline in real-time. Some of the interesting uses for single-pixel systems lie in challenging environments where it is hard to build and maintain conventional optical instruments. The work from the Camera Culture Group aims to help this process by showing ways to augment any lensless or single-pixel system, and suggesting combinations of hardware and computational techniques that can not only enhance existing applications but also enable entirely new ones.