Abstract Ghost imaging (GI) facilitates image acquisition under low-light conditions by single-pixel measurements and thus has great potential in applications in various fields ranging from biomedical imaging to remote sensing. However, GI usually requires a large amount of single-pixel samplings in order to reconstruct a high-resolution image, imposing a practical limit for its applications. Here we propose a far-field super-resolution GI technique that incorporates the physical model for GI image formation into a deep neural network. The resulting hybrid neural network does not need to pre-train on any dataset, and allows the reconstruction of a far-field image with the resolution beyond the diffraction limit. Furthermore, the physical model imposes a constraint to the network output, making it effectively interpretable. We experimentally demonstrate the proposed GI technique by imaging a flying drone, and show that it outperforms some other widespread GI techniques in terms of both spatial resolution and sampling ratio. We believe that this study provides a new framework for GI, and paves a way for its practical applications. 

/pdf/far-field-super-resolution-ghost-imaging-with-a-deep-neural-29c8tu4t.pdf

Far-field super-resolution ghost imaging with a deep neural network constraint

/pdf/far-field-super-resolution-ghost-imaging-with-a-deep-neural-10tu27q0.pdf

Different from conventional imaging methods, which are based on the first-order field correlation, ghost imaging (GI) obtains the image information through high-order mutual-correlation of light fields from two paths with an object appearing in only one path. As a new optical imaging technology, GI not only provides us new capabilities beyond the conventional imaging methods, but also gives out a new viewpoint of imaging physical mechanism. It may be applied to many potential applications, such as remote sensing, snap-shot spectral imaging, thermal X-ray diffraction imaging and imaging through scattering media. In this paper, we reviewed mainly our research work of ghost imaging via sparsity constraints (GISC) and discussed the application and theory prospect of GISC concisely.

https://www.mdpi.com/2076-3417/8/8/1379/pdf

A Review of Ghost Imaging via Sparsity Constraints

Ghost imaging system requires a large number of samples to reconstruct the object. Computational ghost imaging can use well-designed pre-modulated orthogonal patterns to reduce the requirement of sampling number and increase the imaging quality, while the rotating ground glass (RGG) scheme cannot. Instead of the pre-modulation method, a post-processing method using Gram-Schmidt process to orthonormalize the patterns in a RGG scheme is introduced. Reconstructed ghost image after the Gram-Schmidt process (SGI) are tested using the quality indicators such as the Contrast-to-Noise (CNR), the Peak Signal to Noise Ratio (PSNR), the Correlation Coefficient (CC) and reducing the Mean Square Error (MSE). Simulation results show that this method has obvious advantage on enhancing the efficiency of image acquisition, and the sampling number requirement drops from several thousands to a few hundreds in ideal condition. However, in actual system with noise, the image quality from SGI declines in large sampling number, for noise and errors accumulate in the orthonormalization process. So an improved Group SGI method is then developed to avoid this error accumulation, which behaves effectively in reconstructing the image from experimental data and show good performances in large sampling number too. Since this method do not change the relationship between the reference patterns and the bucket values, it can easily combine with most of reconstruction algorithms and improve their reconstruction efficiency.

Orthonormalization method in ghost imaging.

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.

/pdf/sub-nyquist-computational-ghost-imaging-with-deep-learning-10ciiktykz.pdf

Sub-Nyquist computational ghost imaging with deep learning.

Three-dimensional ghost imaging LiDAR via sparsity constraint (3D GISC LiDAR), as a staring imaging method, can obtain both the range information and spatial distribution of a remote target with a single-pixel time-resolved detector. However, previous demonstrations mainly focused on the relatively static scene in visible light. Here we propose an airborne near infrared 3D GISC LiDAR system and airborne high-resolution imaging is implemented. Experimental results show that an image with 0.48 m horizontal resolution as well as 0.5 m range resolution at approximately 1.04 km height can be achieved. Some considerations on the improvement of this LiDAR system are also discussed.

https://www.mdpi.com/2072-4292/10/5/732/pdf

Airborne Near Infrared Three-Dimensional Ghost Imaging LiDAR via Sparsity Constraint

The effect of background light on the imaging quality of three typical ghost imaging (GI) lidar systems (namely narrow pulsed GI lidar, heterodyne GI lidar, and pulse-compression GI lidar via coherent detection) is investigated. By computing the signal-to-noise ratio (SNR) of fluctuation-correlation GI, our analytical results, which are backed up by numerical simulations, demonstrate that pulse-compression GI lidar via coherent detection has the strongest capacity against background light, whereas the reconstruction quality of narrow pulsed GI lidar is the most vulnerable to background light. The relationship between the peak SNR of the reconstruction image and σ (namely, the signal power to background power ratio) for the three GI lidar systems is also presented, and the results accord with the curve of SNR-σ.

Performance analysis of ghost imaging lidar in background light environment

According to the reconstruction feature of fluctuation-correlation ghost imaging (GI), we define a normalized characteristic matrix and the influence of the property of random coded patterns on GI is investigated based on the theory of matrix analysis Both simulative and experimental results demonstrate that for different random coded patterns, the quality of fluctuation-correlation GI can be predicted by some parameters extracted from the normalized characteristic matrix, which suggests its potential application in the optimization of random coded patterns for GI system

https://iopscience.iop.org/article/10.1088/2040-8978/18/6/065703/pdf

Chenglong Wang

Papers

Far-field super-resolution ghost imaging with a deep neural network constraint

Far-field super-resolution ghost imaging with a deep neural network constraint

Airborne Near Infrared Three-Dimensional Ghost Imaging LiDAR via Sparsity Constraint

Performance analysis of ghost imaging lidar in background light environment

The influence of the property of random coded patterns on fluctuation-correlation ghost imaging