An efficient neural network for low sampling computational ghost imaging based on EMNIST training

Recent researches showed that the traditional ghost imaging (TGI) using pseudo-thermal light can be affected by the solid scattering material placed between the object and the light source. This scattering case might also affect the computational ghost imaging (CGI) and is considered in this paper. Different from the TGI system modeled by Huygens–Fresnel theory in continuous coordinates, in CGI the use of computer-generated patterns with the controllable size ( $d$ ) of each pixel discretizes the object space into the $d$ -based pixel-grids. Therefore, how the CGI system is affected by the scatters depends on the distribution of the scattered light on these discretized grids instead of in continuous coordinates, which has not been explored. In this paper, we demonstrate that the “anti-scattering” capacity of CGI is governed by the discrete point-scattering-function (DPSF) that is determined by $\frac{{w_{0}}}{d}$ ( $w_{0}$ : the width of point-scattering-function of the scatters). We show that the blurry effect on the reconstructed image is visible as $d\ll\frac{{w_{0}}}{2}$ , but can be highly restrained as $d$ increasing to be comparable to $\frac{w_{0}}{2}$ or larger. Furthermore, $d$ determines the spatial resolution power of CGI. In the given CGI system, the minimum $d$ of $\sim\frac{{w}_{0}}{5}$ could be achieved with maintaining the behavior of “scattering-free”.

We introduce sinusoidal ghost imaging (SGI), which uses 2D orthogonal sinusoidal patterns instead of random patterns in "computational ghost imaging" (CGI). Simulations and experiments are performed. In comparison with the"differential ghost imaging" algorithm that was used to improve the SNR of ghost imaging, results of SGI show about 3 orders of magnitude higher SNR, which can be reconstructed even with a much smaller number of patterns. More importantly, based on the results, SGI provides the great opportunity to generate innate processed images by predefined selection of patterns. This can speed up detection process considerably and paves the way for real applications.

Demonstration of computational ghost imaging through fog

In the ghost imaging system based on the multidimensional vector matrix Walsh transform, the Walsh speckle pattern is used to continuously sample the moving target object, so there is a certain correlation and complementary detail information between multiple frames in the ghost imaging video. In our previous research, we broke through the inherent limitations of the digital micromirror device refresh rate on ghost imaging systems, allowing us to reconstruct more detailed frames. Therefore, by utilizing the correlation between these detailed frames in the time and space domains, we can improve the comfort of ghost imaging videos from the perspective of improving the single-frame quality of ghost imaging videos. To further improve multi-frame quality by utilizing more detailed frames in ghost imaging videos, this paper proposes a ghost imaging video algorithm based on the multidimensional vector matrix Walsh transform of bidirectional N-aligned fusion frames. Combining deep learning with computational ghost imaging, utilizing a bidirectional N-alignment algorithm and a deep learning neural network framework for target frame integration. The ghost imaging videos obtained from previous research can be improved in terms of noise, motion blur, and single-frame detail richness to enhance the imaging quality of the target frame. This paper constructs an encoding module and a corresponding feature fusion module suitable for multidimensional vector Walsh transform ghost imaging from the perspective of network width and feature multi-branch extraction based on GoogleNet Inception V3. A loss function suitable for four-dimensional vector matrix Walsh transform ghost imaging has been defined, which can better eliminate the noise and distortion caused by Walsh speckle sampling. After comparing the experimental results of moving objects, the results show that the algorithm proposed in this paper has significantly improved structural similarity, blur index, noise index, and other aspects compared to existing ghost imaging video optimization methods. The similarity angle of the NRSS structure has increased by 18.58% compared to the original reconstructed image, the blur parameter has increased by 31.9%, and the noise parameter has risen by 9.22%.

Single-pixel imaging uses a single-pixel detector, rather than a focal plane detector array, to image a scene. It provides advantages for applications such as multi-wavelength, three-dimensional imaging. However, low frame rates have been a major obstacle inhibiting the use of computational ghost imaging technique in wider applications since its invention one decade ago. To address this problem, a computational ghost imaging scheme, which utilizes an LED-based, high-speed illumination module is presented in this work. At 32 × 32 pixel resolution, the proof-of-principle system achieved continuous imaging with 1000 fps frame rate, approximately two orders larger than those of other existing ghost imaging systems. The proposed scheme provides a cost-effective and high-speed imaging technique for dynamic imaging applications.

Ghost imaging has gone through from quantum to classical pseudothermal to computational field over the last two decades. As a kernel part in computational ghost imaging (CGI), the reconstruction algorithm plays a decisive role in imaging quality and system practicality. In order to introduce more prior knowledge into the reconstruction algorithm, existing research adds image patch prior into CGI and improves the imaging efficiency. In this paper, the total variation minimization algorithm via adaptive deep dictionary learning (TVADDL) is proposed to update an adaptive deep dictionary through the CGI reconstruction process. The proposed algorithm framework is able to capture more precise texture features with a multi-layer architecture dictionary and adapt the learned dictionary by gradient descent on CGI reconstruction loss value. The results of simulation and experiment show that TVADDL can achieve higher peak signal-to-noise ratio than the algorithms without patch prior and the algorithms using the shallow dictionary or non-adaptive deep dictionary.

Ghost imaging with single-pixel bucket detector has attracted more and more current attention due to its marked physical characteristics. However, in ghost imaging, a large number of reference intensity patterns are usually required for object reconstruction, hence many applications based on ghost imaging (such as tomography and optical security) may be tedious since heavy storage or transmission is requested. In this paper, we report that the compressed reference intensity patterns can be used for object recovery in computational ghost imaging (with single-pixel bucket detector), and object verification can be further conducted. Only a small portion (such as 2.0% pixels) of each reference intensity pattern is used for object reconstruction, and the recovered object is verified by using nonlinear correlation algorithm. Since statistical characteristic and speckle averaging property are inherent in ghost imaging, sidelobes or multiple peaks can be effectively suppressed or eliminated in the nonlinear correlation outputs when random pixel positions are selected from each reference intensity pattern. Since pixel positions can be randomly selected from each 2D reference intensity pattern (such as total measurements of 20000), a large key space and high flexibility can be generated when the proposed method is applied for authenticationbased cryptography. When compressive sensing is used to recover the object with a small number of measurements, the proposed strategy could still be feasible through further compressing the recorded data (i.e., reference intensity patterns) followed by object verification. It is expected that the proposed method not only compresses the recorded data and facilitates the storage or transmission, but also can build up novel capability (i.e., classical or quantum information verification) for ghost imaging.

Influence of detector response speed on the contrast-to-noise ratio of reflective ghost imaging

We provide a method to evaluate effects of a lossy and noisy optical channel in computational ghost imaging (CGI) technique. Instead of preparing an external noise source, we simulate the optical channel with a basic CGI experiment using programmatically generated noise-induced patterns. By using our method, we show that CGI can reject a noise of which intensity is similar with an imaging signal intensity at a target. The results with our method are well matched with experimental ones including external noise source. This method would provide useful knowledge to analyze environmental effects in CGI without realization of the environment.

We compared images reconstructed by three methods: ghost imaging (GI), Hadamard transform imaging (HTI), and scan-based imaging. Although GI and HTI use a bucket (or single channel) detector, GI has attracted more attention than HTI in recent years. Nevertheless, a direct comparison between them has not yet been conducted to the best of our knowledge. In the present work, we evaluate contrast ratios of images obtained from computational GI (CGI) and HTI under various signal-to-noise ratio (SNR) conditions. Our results indicate that HTI and CGI are useful in high- and low-SNR situations, respectively, although both methods have a similar performance.

A novel optical image hiding scheme based on an expansion strategy is presented under the framework of computational ghost imaging. The image to be hidden is concealed into an expanded interim with the same size as the host image. This is implemented by rearranging the measured intensities of the original object after the process of ghost imaging. An initial Hadamard matrix is used to generate additional matrices by shifting it circularly along the column direction, so that enough 2D patterns are engendered to retrieve phase-only profiles for imaging. Next, the frequency coefficients of the host image are modified with that of the expanded interim by controlling a small weighting factor. After an inverse transform, the host image carrying the hidden information can be obtained with high imperceptibility. Security is assured by considering optical parameters, such as wavelength and axial distance, as secret keys due to their high sensitivity to tiny change. Importantly, differing from other computational ghost imaging based schemes, many phase-only profiles are used to collect the measured intensities to enhance the resistance against noise and occlusion attacks. The simulated experiments illustrate the feasibility and effectiveness of the proposed scheme.

As computational ghost imaging is widely used in the military, radar, and other fields, its security and efficiency became more and more important. In this paper, we propose a compressive ghost imaging encryption scheme based on the hyper-chaotic system, DNA encoding, and KSVD algorithm for the first time. First, a 4-dimensional hyper-chaotic system is used to generate four long pseudorandom sequences and diffuse the sequences with DNA operation to get the phase mask sequence, and then N phase mask matrixes are generated from the sequences. Second, in order to improve the reconstruction efficiency, KSVD algorithm is used to generate dictionary D to sparse the image. The transmission key of the proposed scheme includes the initial values of hyper-chaotic and dictionary D, which has plaintext correlation and big key space. Compared with the existing compressive ghost imaging encryption scheme, the proposed scheme is more sensitive to initial values and more complexity and has smaller transmission key, which makes the encryption scheme more secure, and the reconstruction efficiency is higher too. Simulation results and security analysis demonstrate the good performance of the proposed scheme.

Ghost imaging is a fascinating process, where light interacting with an object is recorded without resolution, but the shape of the object is nevertheless retrieved, thanks to quantum or classical correlations of this interacting light with either a computed or detected random signal. Recently, ghost imaging has been extended to a time object, by using several thousands copies of this periodic object. Here, we present a very simple device, inspired by computational ghost imaging, that allows the retrieval of a single non-reproducible, periodic or non-periodic, temporal signal. The reconstruction is performed by a single shot, spatially multiplexed, measurement of the spatial intensity correlations between computer-generated random images and the images, modulated by a temporal signal, recorded and summed on a chip CMOS camera used with no temporal resolution. Our device allows the reconstruction of either a single temporal signal with monochrome images or wavelength-multiplexed signals with color images.

The non‐local imaging characteristics of computational ghost imaging (CGI) make it widely used in the field of information security. However, the information processing method based on traditional CGI has some problems in the process of multi‐image encryption, such as low efficiency and unsatisfactory reconstruction quality. Therefore, this article proposes to use wavelet transform and singular value decomposition technology to help computational ghost imaging to efficiently complete the coding and encryption of multiple images, and use alternating direction multiplier method (ADMM) in the process of image restoration to carry out high‐quality reconstruction of the original encrypted image. The results show that the scheme has good reconstruction effect, strong security and robustness even at a low sampling rate. It provides some reference for digital watermarking technology and cryptography.

There is a consensus that turbulence-free images cannot be obtained by conventional computational ghost imaging (CGI) because the CGI is only a classic simulation, which does not satisfy the conditions of turbulence-free imaging. In this article, we report a turbulence-immune CGI method based on a multi-scale generative adversarial network (MsGAN). Here, the conventional CGI framework is not changed, but the conventional CGI coincidence measurement algorithm is optimized by an MsGAN. Thus, the satisfactory ghost image can be reconstructed by training the network, and the visual effect can be significantly improved.