MIT Develops Faster Single Pixel Cameras

Technology Dynamics New design principles and optimization algorithms have increased the potential use of lensless imaging systems. Conventional imaging and microscopy techniques use high-power objective lenses to project light reflected from an object onto a suitable sensor, and methods for obtaining high-quality images without the use of lenses have continued to advance. One of the reasons for this is the advancement of modern signal processing technology, which provides an opportunity for imaging technology to shift the key of imaging from optical hardware to computing.

A project at the MIT Media Lab Camera Culture Group has now developed a lensless imaging method that uses compression sensing (CS), a basic numerical method for computational imaging, and current time-resolved optics. Breakthrough in sensor technology. At present, this method is critical for the projects of some research groups.

No lens: Improved single-pixel imaging

The team's research results show that efficient lensless imaging can be achieved through ultra-fast measurement and compression sensing. It also shows that this novel imaging structure can be used when lens imaging is not possible.

MIT’s Guy Satat commented: “As far as we know, this is a time-resolved sensing and single-pixel camera for detecting reflectivity — it is actually the first combination of photographic applications.”

"Single-pixel systems have been well-known and studied for many years, and time-resolved sensing has been used to measure the reflectivity without compression induction and to recover the geometric coordinates of three-dimensional scenes in Lidar (LIDAR), but we have now The key to solving this problem was found, combining these methods to reconstruct the reflectivity and albedo of the scene."

As reported in a paper in the IEEE's "Transactions on Computational Imaging," this new method ultimately results in a 50-fold improvement in compression-induced image acquisition. The lensless single-pixel camera system uses a combined measurement mode using the same sensor pixels under different grating illuminations, where each raster is controlled by a spatial light modulator to encode different information into each measurement - These findings can reduce the number of exposures from the usual thousands to dozens.

Smarter modulation

The MIT research project starts with how to guide designers to achieve the best system structure for lensless applications, and studies three distinct but closely related fundamental issues in lensless imaging. The new design framework created by the research team provides a set of guidelines and decision-making methods, and provides information on how to use the available resources in a given scenario to use compression induction to recover the best image while defining the maximization of the benefits of compressed sensing. .

This does not necessarily involve single-pixel sensing - one of the goals of the framework is to help define when the use of single-pixel systems is most profitable, and when it may not be - but the framework is designed to answer related design issues, for example, Determining the optimal position of a particular sensor, or increasing the time resolution under certain conditions, may be more effective than adding more detectors.

The second idea of ​​this project is to examine the role of time-resolved signals and to clarify how improved temporal resolution reduces the number of individual modulation signals needed to build high-quality images. The third idea is to study how to optimize various modes of modulated light and how to extract more information from each mode.

Satat said: "Compression sensing can adjust light in a more intelligent way. Without it, obtaining high-resolution images can only take a lot of time to make measurements. The geometry of the system will significantly affect the time-resolved measurement because it is closer to the detector. The point will be measured first, and the point farther away will be measured later.The modulated light can avoid this effect and get more information in each measurement, and may only need less modulation modes to get the complete result ."

Optical in complex environments

Time-resolved sensing and femtosecond imaging have already played a role in several projects of Ramesh Raskar, the leader of the camera culture research group, and have shown a variety of practical applications that can be made with accurate measurements of optical transit times.

Significant breakthroughs in this approach include the use of time-resolved sensing to more efficiently image the scattering medium by collecting and evaluating all the photons scattered by the light instead of screening out the most useful ones from the many photons. Another group of researchers studied the method of measuring the bidirectional reflectance distribution function (BRDF), which is a complex parameter related to the reflected light of an object. This method will be an important factor in creating realistic computer graphics and VR animation.

Satat said: “The complete measurement of BRDF is a challenge because it involves a moving source of illumination. The team’s project measured the BRDF by using a streak camera without moving the light source to avoid these difficulties, and we realized that the light The more time is related to the angle of reflection, this allows us to recover more information from the time-resolved signal."

Projects of other research groups include terahertz time-gated spectroscopic imaging for the analysis of hierarchical structures, which may have utility value in industrial inspections, but this technology has long been theoretically proven to be useful for observing things that are currently unknown; It uses open fiber optic bundles to image the locations of traditional single-chip cameras where it cannot be reached, effectively converting the fiber bundles into a set of discrete individual pixels whose spatial positions are not precisely determined.

Outside the Massachusetts Institute of Technology, M Squared and the University of Glasgow in the United Kingdom recently pointed out the potential value of a lensless single-pixel camera system and developed an instrument for real-time imaging of methane gas leaking from a ruptured pipeline. .

Satat said: “Some of the interesting aspects of a single-pixel imaging system is that it can be used in complex environments where it is difficult to build and maintain traditional optical instruments. The purpose of our work in the camera culture research group is to add any The single-pixel system approach helps to implement this process, and combining hardware with computing technology can not only improve the performance of existing solutions, but can even create entirely new solutions."