Investigation Of Signal To Noise Ratio And Application In Single-Pixel Computatlional Imaging

Single-pixel imaging has a long history,which can be traced back to raster scan(RS).In recent years,single-pixel imaging has once again attracted widespread attention from researchers due to the rise of a novel imaging technology named correlated imaging.Correlated imaging technology relies on the correlation between the signal path and the reference path in the imaging system for image reconstruction Two-dimensional or three-dimensional image reconstruction can be performed using a single-pixel detector(SPD)without spatial resolution.Since single-pixel imaging sacrifices time cost in exchange for spatial information,thus the pressure at the detector is greatly reduced.Single-pixel imaging technology is a great alternative technology when the area array detector is difficult to obtain or expensive.Signal to noise ratio(SNR)is widely used in the field of image processing and signal communication for quality evaluation of reconstructed signals.The signal-to-noise ratio directly reflects the proportional relationship between the signal term and the noise term.The magnitude of the signal directly reflects the noise suppression capability of the imaging system By analyzing the signal-to-noise ratio of the reconstructed image in single-pixel imaging,it is helpful to further understand the noise influence of the imaging system,and provide theoretical support as well as a guidance for suppressing noise,it will also make developments for more efficient single-pixel imaging algorithms and systems This thesis first proposes a theory of signal-to-noise ratio analysis of single-pixel imaging,which is verified by experiments.Then,two application scenarios based on theoretical analysis conclusions are further proposedThe main work of this thesis is mainly through the analysis and comparison of the imaging signal-to-noise ratio of computational single-pixel imaging.By constructing a mathematical model,the influence of different kinds of noises on the signal-to-noise ratio is discussed,and the suppression of noise by different single-pixel imaging scanning methods is further analyzed.According to the obtained conclusions,a suitable single-pixel imaging scheme is selected for different scenarios.Multi-pixel structure scanning and point-by-point raster scanning are compared,and the reconstruction performance under the interference of additive noise and multiplicative noise is considered.By using the signal-to-noise ratio as an evaluation tool,the applicable scenarios of the two sampling methods are analyzed and compared.In the case of active illumination with stable light source,compared to the multi-pixel structure sampling that needs to spread the energy to the entire imaging surface,the energy is concentrated to one point for point-by-point scanning,which can maximize the detection signal-to-noise ratio and obtain more signal value.Highly reconstructed signal-to-noise ratio images are therefore more suitable for applications such as remote imaging.For passive illumination with uncontrolled light source,multi-pixel structure sampling can achieve higher weight in the shortest time consumption due to the ability of averaging noise by multi-pixel structure sampling.The comparison work between the two sampling methods can play a reference for the selection of single-pixel imaging scanning methods under different conditions.In the near-infrared region,the imaging quality has been limited by the high cost of the area array detector and the low efficiency of the detector.For near-infrared three-dimensional imaging,it is difficult to obtain high-quality and high-precision three-dimensional reconstruction results using a low-cost imaging system.To this end,based on our theory of noise analysis and comparison work,we use a point-by-point scanning technique to obtain a higher-quality two-dimensional intensity pattern,while using Fourier transform profilometry,only one reconstruction reconstruction pattern is needed.In the case of the height information reconstruction accuracy of the order of millimeters is obtained.Our three-dimensional near-infrared imaging system is simple in structure,easy to implement,and has high reconstruction accuracy.It can be widely used in scenes such as face recognition.Imaging systems that develop high SNR reconstructed images have always been the pursuit of researchers.For the analysis of signal-to-noise ratio,most of the work is to better suppress noise and obtain better reconstruction quality.If you think backwards,consider that you can obtain images with high reconstructed signal-to-noise ratio under normal conditions,and change some parameters to get images with very low signal-to-noise ratio.Such an application scenario and information security encryption work are highly compatible.To this end,based on the previous analysis of signal-to-noise ratio,combined with the conversion between time domain and spatial domain in single-pixel imaging,an information security encryption technology based on computational temporal ghost imaging(CTGI)is proposed.Through the analysis and evaluation of its anti-cutting ability,key scrambling reconstruction effect and partial eavesdropping reconstruction effect,it is found that this encryption security scheme has high security and can be realized by pure digital software or physical optical system.Simple structure and strong robustness are also realized.The proposal of this scheme also opens up a new application field for computational temporal ghost imaging.