Abstract:Objective Computational ghost imaging (CGI) uses a single-pixel detector to realize imaging. It has received a great deal of attention in recent years because it is a low-cost invisible spectrum imaging technology that can make the light transmit via a scattering medium. Researches on CGI through scattering media are mainly conducted on static scattering media such as water and ground glass. In contrast, few studies have been conducted on some complex scattering media, and the imaging situation for dynamic scattering media is unknown. A smoke screen is a type of scattering medium with diffusion and subsidence movements. At the same time, the heat of smoke changes the refractive index of the light. This is a representative of a complex scattering medium. Further, smoke imaging is of research value in the fields of military and life sciences. Dynamic smoke screens distort light intensity measurements and cause image deterioration, which is the difficulty in CGI. In this paper, we choose the smoke screen as the scattering medium for studying the imaging effects under static and dynamic smoke conditions. Smoke screen alleviates dynamically induced imaging degradation and provides a useful reference for CGI in permeable scattering imaging applications. Methods A CGI system with a smoke screen in the detection path was built. A smoke chamber was used as the scattering medium, and a 532-nm laser was used as the light source. Scattering media were classified according to the motion state of smoke particles. The error factors of the dynamic smoke screen environment were analyzed by simulation. Under experimental conditions, imaging results were analyzed using various frame-rate projections. A point-by-point compensation (PPC) method was proposed and used to track the attenuation process of the light intensity caused by smoke motion, with the addition of a "1" measurement matrix before and after each CGI measurement matrix. The projected light intensity of these measurement matrices showed attenuation changes caused by the smoke screen motion. The distortion of light intensity due to the scattering medium was corrected according to the attenuation curve, and the results were compared. Based on analytical and comparative results, the feasibility of the CGI technique was verified, and the applicable range of PPC was obtained. Results and Discussions The moving state of the smoke medium was classified into the static state, slow dynamic state, and fast dynamic state. The imaging results under three types of motion conditions were compared, and the following conclusions were obtained. 1) Simulation results show that CGI has strong robustness in a slow dynamic smoke environment, and the imaging results do not change significantly due to the scattering distortion of the light intensity. However, if smoke motion causes significant change in the light intensity before and after measurements, images will deteriorate significantly. At this moment, the images obtained by the PPC method are sufficiently improved, but the proposed PPC method is also affected by noises, such as ambient light intensity and light source power. If the signal-to-noise ratio of the intensity measurement is less than 60, the CGI and PPC will fail. To improve imaging with PPC, these noises need to be strictly controlled. 2) The CGI system achieves relatively clear images in a static smoke environment, but the traditional imaging method fails. 3) In a slow dynamic smoke environment, the CGI imaging results are not significantly different from that in a static environment. 4) In a fast dynamic smoke screen environment, the lower the frame frequency, the weaker the CGI imaging effect. Conversely, suppose the frame frequency is too high. In that case, it will be difficult for the laser to pass through, making it impossible to image in the initial stages of smoke emission and making the function of CGI uncertain. 5) In a dynamic smoke screen environment, the PPC method can significantly improve the imaging quality when the projected frame frequency is low. This comparisonis are more pronounced at 10 and 50 Hz. Conclusions CGI has obvious anti-scattering imaging capabilities. Imaging results under static and slow dynamic conditions are better, and there is no significant difference. In a fast dynamic smoke environment, CGI cannot work due to the distortion of the measured light intensity, and the lower the frame frequency, the greater the distortion. There is a certain tolerance for the degree of motion of the CGI scattering medium. Suppose the degree of motion is within a tolerance range. In that case, the imaging results show no significant quality reduction or are the same as that in static states, even though the measurements show detectable distortion. The PPC method corrects measurements by tracking the process of light intensity distortion to achieve considerable imaging effects. A powerful and stable laser is a suitable light source for transmitting smoke in a CGI system. In conclusion, CGI has unique advantages in anti-scattering imaging and can be improved by the PPC method using dynamic scattering media.

Computational Ghost Imaging Enhanced by Correlations in Measurements with Different Sampling Ratios

Computational ghost imaging with deep compressed sensing

Computational ghost imaging enhanced by degradation models for under-sampling

Computational Ghost Imaging Using a Dynamic Scattering Medium by the Point-by-Point Compensation Method

Computational Ghost Imaging Via Adaptive Deep Dictionary Learning.

Compressive Sensing Ghost Imaging Based On Neighbor Similarity

Extreme-Quality Computational Imaging via Degradation Framework

Computational ghost imaging using deep learning

High-speed computational ghost imaging with compressed sensing based on a convolutional neural network

Compressive Sensing Ghost Imaging Object Detection Using Generative Adversarial Networks

Compressive sensing ghost imaging based on image gradient

Mean Square Convergence Unbiased Estimation of Thermal Light Correlated Imaging

Experimental Study on Laser Computational Ghost Imaging Through Smoke Media

Computational Ghost Imaging in Scattering Media Using Simulation-Based Deep Learning

A Novel Approach of Parallel Retina-Like Computational Ghost Imaging

Parallel computational ghost imaging with modulation patterns multiplexing and permutation inspired by compound eyes

Anti-scattering medium computational ghost imaging with modified Hadamard patterns

Deep-learning denoising computational ghost imaging

Computational ghost imaging using the dilated ghost network

High Speed Computational Ghost Imaging via Spatial Sweeping

Transformer-based flexible sampling ratio compressed ghost imaging