Computational Ghost Imaging Research Articles

X-ray imaging of fast dynamics with single-pixel detector.

We demonstrate experimentally the ability to use a single-pixel detector for two-dimensional high-resolution x-ray imaging of fast dynamics. We image the rotation of a spinning chopper at 100 kHz and at spatial resolution of about 40 microns by using the computational ghost imaging approach. The technique we develop can be used for the imaging of fast dynamics of periodic and periodically stimulated effects with a large field of view and at low dose.

Ultra-fast vivid computational ghost imaging of still and moving objects by sweeping random patterns

A sweeping algorithm is introduced in conventional computational ghost-imaging as a method for improving image quality and reducing the shot number. In this method, the intensity of rows and columns in the generated random patterns are set as bright pixels. Since this sweeping is repeated one by one on each row and column, a constant shot number, equal to the sum of raw and column number, is pre-defined. As a result, application of this method to ghost imaging of moving objects offers lower shot numbers and leads to higher imaging speeds. It is shown that the final image can be retrieved with considerable improvements in its visibility, signal to noise and contrast to noise ratios. For still objects, enhancements up to 27, 5 and 57.5 are achieved for the visibility, signal to noise and contrast to noise ratios, respectively. As for the moving objects, we report a maximum imaging speed of 2.79 mm s−1 which is around 4000 times the previously reported data.

Comprehensive comparison of single-pixel imaging methods

Single-pixel imaging is a novel computational imaging scheme. It allows one to use a spatially unresolvable photodetector to acquire the spatial information of objects and reconstruct images by a computational means. The core of single-pixel imaging is encoding the spatial information of objects into one-dimensional light signals by using spatial light modulation and subsequently decoding the information to reconstruct the image. Such a novel imaging scheme has a potential of tacking some challenges in conventional imaging, such as imaging with non-visible light, imaging under weak-light conditions, imaging through turbid media, etc. However, the practicality of single-pixel imaging is limited by its imaging quality and efficiency. Thus, how to improve the imaging quality and reduce the imaging time in single-pixel imaging is extensively explored in the field. In this paper, we comprehensively investigate five representative and widely used single-pixel imaging methods – computational ghost imaging, compressive sensing ghost imaging, Hadamard single-pixel imaging, Fourier single-pixel imaging, and binary Fourier single-pixel imaging. We review the principle of these methods and compare the performance of these methods by numerical simulations and experiments. The comparison not only reveals the connections and differences among these methods, but also shows the advantages and disadvantages of their own. This paper also discusses the problems that remain to be solved in single-pixel imaging and prospects for the future development.

Computational Ghost Imaging with Multiplexed Time-Varying Signals

This study proposes two methods of optical watermarking based on multiplexed time-varying signals for computational ghost imaging using the Hadamard matrices. The proposed methods can realize image fusion and dual optical encryption. The time-varying signal is encoded into a specific Hadamard coefficient in advance and hidden in the light source of the transmitting end as a multiplicative factor or loaded at the receiving end as an additive factor. Theory and experiments confirm the feasibility of this scheme. Moreover, the scheme is highly scalable and has potential applications in multispectral single-pixel imaging.

Forgery attack on optical encryption based on computational ghost imaging.

Attack techniques on a cryptosystem include not only cryptanalysis, but also forgery and modification of messages, deception and confusion on both sender and receiver sides, and so on. In this Letter, we show that an optical encryption system based on computational ghost imaging (CGI) has security vulnerability owing to its high tolerance for error deviation of ciphertext. It leaves a chance for a forgery attack in which attackers can forge a set of fake keys according to the intercepted ciphertext. If the forged key can be transmitted to the receiver by some disguised means, he/she may be cheated or confused by the retrieved fake images. The discovery of this vulnerability may also help upgrade the CGI-based encryption system.

DeepGhost: real-time computational ghost imaging via deep learning

The potential of random pattern based computational ghost imaging (CGI) for real-time applications has been offset by its long image reconstruction time and inefficient reconstruction of complex diverse scenes. To overcome these problems, we propose a fast image reconstruction framework for CGI, called “DeepGhost”, using deep convolutional autoencoder network to achieve real-time imaging at very low sampling rates (10–20%). By transferring prior-knowledge from STL-10 dataset to physical-data driven network, the proposed framework can reconstruct complex unseen targets with high accuracy. The experimental results show that the proposed method outperforms existing deep learning and state-of-the-art compressed sensing methods used for ghost imaging under similar conditions. The proposed method employs deep architecture with fast computation, and tackles the shortcomings of existing schemes i.e., inappropriate architecture, training on limited data under controlled settings, and employing shallow network for fast computation.

Multiple layer encryption and steganography via multi-channel ghost imaging

Multiple layers of information security are introduced based on computational ghost imaging (CGI). We show, in the first step, that it is possible to design a very reliable image encryption scheme using 3D computational ghost imaging with two single-pixel detectors sending their data through two channels. It is then shown that further level of security can be achieved by merging data carrying channels into one and using a coded order for their placement in the sequence of bucket data carried by the single channel. Through the Normalized Root Mean Square scale, we show that the second security level is a one step forward. Yet another layer of security is introduced through hiding the actual grayscale image (or even a color image) inside another image (object) before the ghost imaging process such that the hidden image cannot be recognized by the naked eye. We then successfully retrieve the hidden image from a CGI-reconstructed image despite the presence of noise while such noises make steganography impossible by conventional methods. We show that CGI-based steganography is intrinsically robust to attacks involving manipulation of the image matrix and pixels. As for the cryptography attacks, it is shown that the proposed scheme increases the security and robustness such that an attacker needs more than 96 percent of the coded order to recover the hidden image. Storing a gray scale image in a ghost image and retrieving different intensities for the hidden image is unprecedented and could be of interest to the information security community. We also present performance, robustness and attack analyses and show that our layered configuration for information encryption based on CGI successfully passes these checks.

Deep-learning denoising computational ghost imaging

We propose a deep learning denoising computational ghost imaging (CGI) method to obtain a clear object image with a sub-Nyquist sampling ratio. We develop an end-to-end deep neural network (DDANet) for CGI image reconstruction. DDANet uses a one-dimensional (1-D) bucket signals (BSs) and multiple tunable noise-level maps as input, and outputs a clear image. We train DDANet with simulated BSs and ground-truth pairs, and then retrieve the object image directly from an experimental obtained 1-D BSs. The effectiveness of the proposed method is experimentally investigated. The proposed method has practical applications in image denoising and enhancement of the CGI and single-pixel computational imaging.

Ghost spintronic THz-emitter-array microscope

Terahertz (THz) waves show great potential in nondestructive testing, biodetection and cancer imaging. Despite recent progress in THz wave near-field probes/apertures enabling raster scanning of an object’s surface, an efficient, nonscanning, noninvasive, deep subdiffraction imaging technique remains challenging. Here, we demonstrate THz near-field microscopy using a reconfigurable spintronic THz emitter array (STEA) based on the computational ghost imaging principle. By illuminating an object with the reconfigurable STEA followed by computing the correlation, we can reconstruct an image of the object with deep subdiffraction resolution. By applying an external magnetic field, in-line polarization rotation of the THz wave is realized, making the fused image contrast polarization-free. Time-of-flight (TOF) measurements of coherent THz pulses further enable objects at different distances or depths to be resolved. The demonstrated ghost spintronic THz-emitter-array microscope (GHOSTEAM) is a radically novel imaging tool for THz near-field imaging, opening paradigm-shifting opportunities for nonintrusive label-free bioimaging in a broadband frequency range from 0.1 to 30 THz (namely, 3.3–1000 cm−1).

Spectral encoded computational ghost imaging

Multispectral imaging can acquire images of a scene in multiple spectral bands and has been widely applied in many fields, such as satellite remote sensing and geological exploration. Traditional multispectral imaging employs dispersive optical devices or narrowband filters to separate different wavelengths of light and then uses an array detector or many single-pixel detectors to retrieve them separately. These techniques suffer from large data volumes and low efficiency. In this paper, a spectral encoded strategy is proposed to realize multispectral imaging with computational ghost imaging technology. The projected illumination patterns are produced by means of deterministic orthogonal Hadamard basis patterns fused with random spectral encoded matrices, which represent different spectral information. The evolutionary compressive sensing technique is employed to restore the fused multispectral image of the color object of interest, and then a compressive sensing algorithm is applied to restore the spectral channel images. The validity of the technique is verified by a numerical simulation and an experiment. The proposed method allows a multispectral image simultaneously by only one single-pixel detector, which simplifies the structure of the spectral imaging setup and reduces the amount of data acquired and reconstruction time, and thus improving multispectral imaging efficiency.

Hybrid cryptosystem based on plaintext related computational ghost imaging encryption and elliptic curve algorithm

A hybrid cryptosystem based on plaintext-related ghost imaging and elliptic curve public-key cryptographic algorithm (ECC) is proposed in this paper. Thereinto, the ECC algorithm is used to protect and distribute session keys and the optical encryption system with session keys is used to encrypt and decrypt images. Thousands of phase-only mask keys can be replaced by parameter keys in this method, which successfully reduces the pressure of key storage. With dynamic updating session keys, the attacks are effectively prevented and the security resilience is enhanced. In addition, the utilization of an ECC cryptosystem solves the inevitable problem of key management and dispatch in conventional ghost imaging encryption system. The feasibility, security and robustness of the method are further analyzed through computer simulations.

High-resolution sub-sampling incoherent x-ray imaging with a single-pixel detector

X-ray “ghost” imaging has drawn great attention for its potential to obtain images with a high resolution and lower radiation dose in medical diagnosis, even with only a single-pixel detector. However, it is hard to realize with a portable x-ray source due to its low flux. Here, we demonstrate a computational x-ray ghost imaging scheme where a real bucket detector and specially designed high-efficiency modulation masks are used, together with a robust deep learning algorithm in which a compressed set of Hadamard matrices is incorporated into a multi-level wavelet convolutional neural network. With a portable incoherent x-ray source of ∼37 µm diameter, we have obtained an image of a real object from only 18.75% of the Nyquist sampling rate. A high imaging resolution of ∼10 µm has been achieved, which is required for cancer detection and so represents a concrete step toward the realization of a practical low cost x-ray ghost imaging camera for applications in biomedicine, archeology, material science, and so forth.

Computational ghost imaging system with 4-connected-region-optimized Hadamard pattern sequence

We propose a computational ghost imaging system based on the Hadamard patterns. A 4-connected-region-based method is proposed to optimize the Hadamard pattern sequence and a differential decomposition method is developed to construct a sequence that can be shown on the light modulation devices with the optimized pattern sequence. Finally, the image is reconstructed by the second-order correlation algorithm. Both numerical model and experimental setup are established and a series of experiments are implemented. The imaging performance of the proposed method is validated by comparing with the results from the default order of the Hadamard pattern sequence, Russian-dolls ordering and fast Walsh-Hadamard transform. The proposed system can realize high quality imaging with a low sampling ratio (5%–10%), and it requires much less measurement numbers for imaging.

Deep ghost phase imaging

Deep-learning-based single-pixel phase imaging is proposed. The method, termed deep ghost phase imaging (DGPI), succeeds the advantages of computational ghost imaging, i.e., has the phase imaging quality with high signal-to-noise ratio derived from the Fellgett's multiplex advantage and the point-like detection of diffracted light from objects. A deep convolutional neural network is learned to output a desired phase distribution from an input of a defocused intensity distribution reconstructed by the single-pixel imaging theory. Compared to the conventional interferometric and transport-of-intensity approaches to single-pixel phase imaging, the DGPI requires neither additional intensity measurements nor explicit approximations. The effects of defocus distance and light level are investigated by numerical simulation and an optical experiment confirms the feasibility of the DGPI.

Optical ghost cryptography and steganography

As an indirect imaging technique, computational ghost imaging (GI) obtains the object information by calculating the intensity correlation between a series of computer-generated matrices and the corresponding bucket signals, which thereby offers a potential application in optical encryption. Here, we propose a new steganography scheme, called ghost steganography, based on the principle of computational GI. In our ghost steganography scheme, the bucket intensity signals of a secret image are concealed into the ones of a non-secret image by applying a non-conspicuous number integration process. To further increase the security, we introduce RSA cryptography to encode the integrated bucket signals after the steganography process. Simulation and experiment results fully demonstrate the feasibility of our optical ghost cryptography and steganography scheme. Our work paves a way to the application of GI in steganography and also enriches the knowledge of symmetric and asymmetric optical cryptography. @ Elsevier, 2020.

Computational ghost imaging with discrete stochastic sources

Lots of digital devices were widely used in computational ghost imaging (CGI) to serve as optical sources and detectors. These digital devices greatly affect the statistical behaviors of the CGI system. In this paper, we attribute all the influence of these digital devices to the discrete probability density functions of the optical sources. We investigate CGI with three discrete optical sources, which emit intensities fulfilling Bernoulli, binomial, and Poisson distributions. We clarify the inner relation between the bucket signals and the reference signals by using their joint density probability functions. In correlation measurement, the visibility and signal-to-noise ratio of the ghost images greatly depend on the probability statistics of the sources. We find that CGI with discrete sources can surpass the classical visibility limit one-third in the thermal-light case. The simulation results confirm our theoretical predictions. Thus, the utility of the digital devices makes CGI manipulable and applicable in practice.

Plaintext-related multiple-image encryption based on computational ghost imaging

ABSTRACTThe traditional ghost imaging encryption system has high pressure in key transmission and management for involving a series of random phase-only masks (POMs) as keys. And conventional ghost imaging encoding is vulnerable to a chosen-plaintext attack owing to the linearity of the system. To address these problems, a plaintext-related ghost imaging encryption system which can reduce the key size and ensure the security of the system was proposed in this study. It uses parameter keys and plaintext information to generate all random POMs for encryption of plaintext. POMs used to encrypt images depend on both parameter keys and plain image, thus protecting information from cracking by the chosen-plaintext. At the same time, a multiple-image encryption method based on plaintext-related ghost imaging encryption was developed to improve the encryption efficiency of the system. The proposed method is the characteristic of strong multiple-image encryption ability, high security and enough simple structure.

Sub-Nyquist computational ghost imaging with deep learning.

We propose a deep learning computational ghost imaging (CGI) scheme to achieve sub-Nyquist and high-quality image reconstruction. Unlike the second-order-correlation CGI and compressive-sensing CGI, which use lots of illumination patterns and a one-dimensional (1-D) light intensity sequence (LIS) for image reconstruction, a deep neural network (DAttNet) is proposed to restore the target image only using the 1-D LIS. The DAttNet is trained with simulation data and retrieves the target image from experimental data. The experimental results indicate that the proposed scheme can provide high-quality images with a sub-Nyquist sampling ratio and performs better than the conventional and compressive-sensing CGI methods in sub-Nyquist sampling ratio conditions (e.g., 5.45%). The proposed scheme has potential practical applications in underwater, real-time and dynamic CGI.

Online adaptive computational ghost imaging

We propose an online adaptive computational ghost imaging (CGI) scheme that offers high quality imaging with few measurement numbers. Different from the conventional second-order correlation CGI, the proposed scheme utilizes an online imaging method for image reconstruction, where the object image is recovered immediately after the light intensity is acquired. An adaptive measurement termination method is proposed to terminate the online imaging when the quality of the recovered image is satisfied. The effectiveness of the proposed scheme is verified by numerical simulations and experiments and its imaging performance is assessed by comparing the results with state-of-the-art FWHTGI (fast Walsh-Hadamard transform CGI) and RDCGI (Russian dolls ordering CGI). The results demonstrate that the proposed scheme can achieve high-quality imaging with few measurements, and imaging quality is better than that with FWHTGI and RDCGI under low measurement number conditions. The proposed scheme may be helpful for high-resolution and real-time CGI.

Influence of the source’s energy fluctuation on computational ghost imaging and effective correction approaches

We investigate the influence of the source’s energy fluctuation on both computational ghost imaging and computational ghost imaging via sparsity constraint, and if the reconstruction quality will decrease with the increase of the source’s energy fluctuation. In order to overcome the problem of image degradation, a correction approach against the source’s energy fluctuation is proposed by recording the source’s fluctuation with a monitor before modulation and correcting the echo signal or the intensity of computed reference light field with the data recorded by the monitor. Both the numerical simulation and experimental results demonstrate that computational ghost imaging via sparsity constraint can be enhanced by correcting the echo signal or the intensity of computed reference light field, while only correcting the echo signal is valid for computational ghost imaging.