The use of quantum properties of light has paved new avenues in the field of imaging. Quantum illumination and ghost-imaging protocols have shown exciting promise in addressing the issues inherent in classical imaging, such as background noise and a high level of required illumination. Moreover, quantum metrological schemes analyzing the modal content of light would address the resolution limits of direct imaging, promising enhancement beyond the Rayleigh limit. Augmenting these novel techniques with machine learning, particularly deep learning architectures such as Convolutional Neural Networks (CNNs), one strives to see significant improvements in image reconstruction and object identification in quantum imaging protocols. In my talk, I present recent progress and development on reconstructing the image of different objects with quantum sources, employing a high degree of spatial correlation between photon pairs and autoencoder deep learning architecture.
The use of quantum properties of light has paved new avenues in the field of imaging. Quantum illumination and ghost-imaging protocols have shown exciting promise in addressing the issues inherent in classical imaging, such as background noise and a high level of required illumination. Moreover, quantum metrological schemes analyzing the modal content of light would address the resolution limits of direct imaging, promising enhancement beyond the Rayleigh limit. Augmenting these novel techniques with machine learning, particularly deep learning architectures such as Convolutional Neural Networks (CNNs), one strives to see significant improvements in image reconstruction and object identification in quantum imaging protocols. In my talk, I present recent progress and development on reconstructing the image of different objects with quantum sources, employing a high degree of spatial correlation between photon pairs and autoencoder deep learning architecture.
Quantum imaging using photon pairs with strong quantum correlations has been harnessed to bring quantum advantages to various fields from biological imaging to range finding. Such inherent non-classical properties support the extraction of more valid signals to build photon-limited images, even in low-light conditions where the shot noise becomes dominant as light decreases to a single-photon level. Numerical optimization algorithms are possible but require thousands of photon-sparse frames, and they are thus unavailable in real time. We demonstrate fast correlated-photon imaging enhanced by deep learning as an intelligent computational strategy to discover a deeper structure in big data. Our work verifies that a convolutional neural network can efficiently solve inverse imaging problems associated with strong shot noise and background noise (electronic noise, scattered light). Our results show that we can overcome limitations due to the trade-off between imaging speed and image quality by pushing the low-light imaging technique to the single-photon level in real time, which enables deep-learning-enhanced quantum imaging for real-life applications.
The UK is continuing its drive to commercialise Quantum Technology in Imaging, Sensing, Communications and Computing. In the first part of my talk I will give an overview of the work of QuantIC, the UK hub tasked with bringing quantum technology to bear on real world imaging challenges, across the industrial, scientific, security and consumer markets.
In the second part of my talk I will feature some of my own group’s work on using quantum illumination to both increase image resolution and suppress background and detector noise. Pairs of photons that are quantum entangled with each other are correlated in many degrees of freedom, creation-time, energy, lateral position etc. Fundamentally this means that measuring one of the photons gives near-perfect prediction of the corresponding property of the other photon. In an imaging context, this second photon with known properties can then be used to probe an object. Using the correlations inherent in this entanglement we are able to increase the resolution of an imaging system and suppress both sensor noise and background light. Our immediate goal is to recover an image of the object without interference from any thermal background or other light sources that might otherwise obscure or confuse what we are seeing. This work in the optical regime has strong parallels with proposals around quantum radar and it is an interesting exercise to consider what one might inform with respect to the other.
The technologies utilizing quantum states of light have been in the spotlight for the last two decades. In this regard, quantum metrology, quantum imaging, quantum-optical communication are some of the important applications that exploit fascinating quantum properties like quantum superposition, quantum correlations, and nonclassical photon statistics. However, the state-of-art technologies operating at the single-photon level are not robust enough to truly realize a reliable quantum-photonic technology. In Chapter 1, I present a historical account of photon-based technologies. Furthermore, I discuss recent efforts and encouraging developments in the field of quantum-photonic technologies, and major challenges for the experimental realization of reliable quantum technologies utilizing photons, setting up a stage for unveiling our smart methodologies to cope with them. Similarly, in Chapter 2, I review the fundamental concepts such as states of light, spatial modes of light, and machine learning. In Chapter 3 of this dissertation, I present a theoretical work on a nonlinear quantum metrology scheme, showing a sub-shot-noise limited phase estimation using the displaced-squeezed light and on/off detection. Furthermore, I discuss a camera-based squeezed-light detection technique that can be a smart and time-efficient alternative to conventional balanced-homodyne detection. In Chapter 4, I discuss our efforts to incorporate artificial intelligence in a quest to improve the efficiency of discriminating thermal light sources from coherent light sources. The conventional identification technique requires a large number of measurements. We utilize artificial neural networks to dramatically reduce the number of measurements required to distinguish thermal light and coherent light. In Chapter 5, I present a communication protocol that utilizes the spatial modes of light. Despite being valuable resources for a wide variety of quantum technologies, spatial modes of light are extremely vulnerable to random phase fluctuations. The conventional techniques to cope with these challenges are relatively inefficient. We utilize convolutional neural networks to perform the spatial mode correction of single photons, resulting in a near-unity fidelity of correction. I wrap up my dissertation in Chapter 6 by summarizing the historical context of quantum technologies, challenging problems facing state-of-art quantum technologies, and the importance of our efforts to introduce artificial intelligence in photonic technologies.
Correlated photon pairs, carrying strong quantum correlations, have been harnessed to bring quantum advantages to various fields from biological imaging to range finding. Such inherent non-classical properties support extracting more valid signals to build photon-limited images even in low flux-level, where the shot noise becomes dominant as light source decreases to single-photon level. Optimization by numerical reconstruction algorithms is possible but require thousands of photon-sparse frames, thus unavailable in real time. Here, we present an experimental fast correlated-photon imaging enhanced by deep learning, showing an intelligent computational strategy to discover deeper structure in big data. Convolutional neural network is found being able to efficiently solve image inverse problems associated with strong shot noise and background noise (electronic noise, scattered light). Our results fill the key gap in incompatibility between imaging speed and image quality by pushing low-light imaging technique to the regime of real-time and single-photon level, opening up an avenue to deep leaning-enhanced quantum imaging for real-life applications.
The process of image recording is arguably one of the most prevalent technology in modern society and continues to inspire vast swathes of research due to its widespread applications spanning military, medical and consumer spheres. The danger present in a field so broad is that separate niches of research can become isolated with critical advancements struggling to traverse the gulfs between. Unifying the field is the omnipresent drive to acquire an ever increasing quality of images at the lowest possible cost, a goal which warrants continual fundamental research. Quantum entanglement exhibits many intriguing characteristics which make it a suitable tool for such fundamental investigations. The process of spontaneous parametric down- conversion has offered a yet unbeaten strength of photon correlations with the quantum nature of their production providing a reliable and controllable source of single-photons. Arising from these attributes is the technique known as ghost imaging. Though now known to be classically possible, the strength of entanglement generated correlations is yet to be surpassed. This thesis implemented this technique in tandem with the cutting edge detector technology in order to probe the fundamentals of image formation. This form of imaging allowed us to subject an object to a known number of photons whilst acquiring structural information from spatially separate, correlated photons which never interact with the object. The strength of the produced correlations allow us to acquire low background, high resolution images with far less light than traditional techniques and affords many novel benefits. The possibility of incorporating this technique with pre-existent regimes allowed me to draw from advancements made across the landscape of imaging research. Although referred to as “quantum ghost imaging” throughout this work, it should be noted that the intrinsic quantum nature of the correlations was not directly relied upon but provided an ideal source of strongly correlated photons. In order to determine the limits of a traditional imaging system this thesis first sought to answer the question: “can an image of an object be reconstructed from fewer photons than ii pixels in the image?” In chapter 3 I approached this from the perspective of compression, which minimises redundant information within a signal. This lead to the development of an imaging regime capable of imaging with far fewer photons than pixels in the image. By employing assumptions about the sparsity of natural images I was able to reconstruct an image of a biological sample containing an average of less than one photon per image pixel. Having reduced the number of photons necessary to form an image I then considered alternative methods for reducing the optical energy impinging on a sample. I sought to answer the question: “can non-degenerate ghost imaging reduce the optical energy impinged upon an object during imaging”. The photons produced in SPDC need not be of similar wavelengths, however may be chosen far from degeneracy, i.e. non-degenerate. In Chapter 4 I presented a ghost imaging system which illuminated the object with infrared light whilst recording the structural information via entangled visible photons. This allowed for objects opaque to visible light to be imaged in high quality without the need for a spatially resolving infrared detector, the state of the art of which lags behind their silicon based visible counterparts. I presented the systems capabilities by imaging objects which were etched into a gold substrate layered on to silicon, both of which are opaque to visible light. Not only did a reduction in energy deposition arise from the lower energy probe wavelength but applying the reconstruction techniques from the previous chapter brought that down to as low as ≈16 nJcm−2s−1. Seeking to expand the repertoire of applications, the low-light capabilities of my ghost imaging were applied to the technique of phase-contrast microscopy in chapter 5. Typically applied to translucent objects, phase-contrast imaging transfers phase information, i.e. the refractive index changed within the object, into an intensity distribution through the use of a phase-filter. In many of these applications the objects tend to be biological in nature, where high optical exposure can result in bleaching or damage. By applying the phase-filter non- locally, i.e. to the photons correlated to those probing the object, I acquired edge-enhanced images of a phase object whilst illuminating with significantly fewer photons than standard phase-contrast techniques. Having displayed the broad applicability of our low-light ghost imaging system, I then sought to determine the optical resolution in chapter 6. The resolution limits of ghost imaging are not clear at first glance owing to the resolutions dependence upon the strength of spatial correlations. As the length over which the spatial correlations are produced can be brought below the standard diffraction limit, it would seem the resolution of the system could be brought similarly low. To clarify this I artificially restricted the number of spatial modes in each of the correlated beams to uncover the physically realisable resolution. I show that although the resolution of a ghost imaging system to be fundamentally determined by the strength of the correlations, this can never be reached due to the inherent limitations of the intervening imaging system.
Quantum correlation, as an intrinsic property of quantum mechanics, has been widely employed to test the fundamental physical principles and explore the quantum-enhanced technologies. However, such correlation would be drowned and even destroyed in the conditions of high levels of loss and noise, which drops into the classical realm and renders quantum advantage ineffective. Especially in low light conditions, conventional linear classifiers are unable to extract and distinguish quantum and classical correlations with high accuracy. Here we experimentally demonstrate the classification of quantum correlation using deep learning to meet the challenge in the quantum imaging scheme. We design the convolutional neural network to learn and classify the correlated photons efficiently with only 0.1 signal photons per pixel. We show that decreasing signal intensity further weakens the correlation and makes an accurate linear classification impossible, while the deep learning method has a strong robustness of such task with the accuracy of 99.99%. These results open up a new perspective to optimize the quantum correlation in low light conditions, representing a step towards diverse applications in quantum-enhanced measurement scenarios, such as super-resolution microscope, quantum illumination, etc.