Light transport and imaging through complex media

Atmospheric lidar observations provide a unique capability to directly observe the vertical column of cloud and aerosol scattering properties. Detector and solar-background noise, however, hinder the ability of lidar systems to provide reliable backscatter and extinction cross-section estimates. Standard methods for solving this inverse problem are most effective with high signal-to-noise ratio observations that are only available at low resolution in uniform scenes. In this talk, I will describe novel methods for solving the inverse problem with high-resolution, lower signal-to-noise ratio observations that are effective in non-uniform scenes. In particular, we will examine a regularized maximum likelihood formulation of the reconstruction problem, where the regularizer is based on state-of-the-art patch-based imaging denoising methods like BM3D. This regularizer is nonconvex, so care must be taken when initializing any iterative optimization method. We develop a novel coarse-to-fine proximal gradient optimization algorithm in which we step across different levels of image resolution to compute successively better initial points for the proposed optimization procedure. Two case studies of real experimental high spectral resolution lidar data illustrate the advantages associated with the proposed method over the standard approach.

The use of quantum states brings information technologies to a principally new level. At the same time, quantum information is difficult to deal with due to its fragility to detection losses and noise. A remedy is phase sensitive amplification, proposed theoretically more than two decades ago but applied to experiments only recently. 

In particular, phase measurement below the shot-noise level, possible with squeezed light, is strongly affected by detection losses. At the same time, amplification of the quadrature containing the phase information enables overcoming any level of loss. This tolerance to loss has been demonstrated in a recent experiment with two coherently pumped high-gain parametric amplifiers, the first of them producing squeezed light testing the phase and the second one amplifying and protecting the phase information. 

The same principle of phase sensitive amplification can be used to protect from loss the protocol of sub-shot-noise imaging, in which an object is placed into one of twin beams and the image is restored in the difference intensity distribution. By amplifying the image before detection, the limitations imposed by losses can be lifted, and the protocol can be extended to ‘difficult’ spectral ranges such as infrared. 

Time of Flight Non-Line-of-Sight (NLOS) imaging has been used to reconstruct scenes that are blocked from direct view via non-specular reflections in the scene. Current methods image scenes of few meter diameter at centimetre resolution There are currently several challenges in this area. I this talk I will review the current state of our research and the field and point out some of the challenges and opportunities of this method.

Including these challenges are speed and efficiency of the reconstruction; scene features like occlusions and multiple reflections, that cannot be handled by current reconstruction methods; and the development of hardware systems that can capture relevant information at reasonable speed and resolution while maintaining acceptable requirements to size, weight, power, and cost. 

We will highlight several recent developments that seek to address these problems, talk about fundamental limitations and requirements of NLOS imaging , and speculate on application areas and scenarios where this method addresses problems that cannot be better addressed by other competing options, such as non-line-of-sight sonar and RADAR imaging through walls.

Optogenetics has revolutionized neuroscience by enabling remote activation or inhibition of specific populations of neurons in intact brain preparations through genetically targeted light sensitive channels and pump. Nevertheless, studying the role of individual neurons within neuronal circuits is still a challenge and requires joint progresses in opsin engineering and light sculpting methods. 

Here we show that computer generated holography using an amplified pulse laser combined with high light sensitive and somatic opsins enable precise in vitro and in vivo control of neuronal firing in mouse brain with millisecond temporal precision, single cell resolution and unprecedented low illumination level. 

We also show a new optical system generates multiple extended excitation spots in a large volume with micrometric lateral and axial resolution. Two-dimensional temporally focused shapes are multiplexed several times over selected positions, thanks to the precise spatial phase modulation of the pulsed beam. This permits, under multiple configurations, the generation of tens of axially confined spots in an extended volume, spanning a range in depth of up to 500 m. We demonstrate the potential of the approach by performing multi-cell volumetric excitation of photoactivatable GCaMP in the central nervous system of Drosophila larvae, a challenging structure with densely arrayed and small diameter neurons, and by photoconverting the fluorescent protein Kaede in zebrafish larve. Our technique paves the way for the optogenetic manipulation of a large number of neurons in intact circuits.

Cameras are often marketed in terms of the number of pixels they have – the more pixels the “better” the camera.  Rather than increasing the number of pixels we ask the question “how can a camera work when it only has a single pixel?” This talk will link the field of computational ghost imaging to that of single-pixel cameras explaining how components found within a standard data projector, more commonly used for projecting films and the like, can be used to create both still and video cameras using a single photodiode.
These single pixel approaches are particularly useful for imaging at wavelengths where detector arrays are either very expensive or even unobtainable. The ability to image at unusual wavelengths means that one can make cameras that can see through fog or smoke or even image invisible gases as they leak from pipes.
Beyond imaging at these unusual wavelengths, by using pulsed illumination and adding time resolution to the camera it is possible to use a single-pixel camera to see in 3-dimensions, perhaps useful for autonomous vehicles and other robotic applications. 

When monochromatic light propagates through a scattering medium, it is scrambled and produces a seemingly random speckle pattern. This randomization process prevents information to pass through a turbid material, which behave like a screen and prevents us to see through it. Yet, multiple scattering is not enough to completely decouple the two sides of a turbid medium, as interference between the scattered waves is known to produce correlations in the speckle patterns. In most cases these correlations can be used to extract information about the turbid medium itself. An exception is the optical memory effect, that connects the incident with the transmitted wavefront, and thus can be exploited to retrieve information through an otherwise opaque screen. Here we present the experimental characterization of a novel form of correlation that connects the reflected and the transmitted speckle measuring only the reflected light. We characterize the rich phenomenology of this correlation, and show that it can be used to image non-invasively through a strongly scattering medium.

Coherent light transport through disordered media gives rise to the well known speckle pattern. This pattern can be understood as the result of of a complex mixing of the incident coherent light through linear elastic multiple scattering, i.e. the propagation can be described by a transmission matrix, linking input to output modes. I will show how the knowledge of the transmission matrix allows imaging through complex media, and how we can extend these concepts in the realms of computational imaging and even optical computing.