Research On The Key Problems Of Image Quality Improvement In Correlation Imaging Systems

Correlation imaging is a novel imaging modality which extracts the object's information from the second-order intensity or fluctuation correlation of the light fields.This counter-intuitive indirect imaging technique divides the light source into the two different optical paths:in the signal arm,a point-like bucket detector points directly to the object,and collects the total intensity reflected or transmitted from the object;while in the reference arm,the light field is freely propagating through the optical system,and the spatial distribution of the light fields are recorded and spatially-resolved by a scanning device or a camera.The image of the object would then be reconstructed by correlating the responses of the two separate detectors in the different optical paths.Since the first observation of this phenomenon in the 1990s,correlation imaging has been investigated by the community for nearly three decades,and the research interests are ranging from the exploration of the physical mechanism,to the development of new schemes or new methods for reconstructing fine images,including the applications of correlation imaging to a variety of challenging imaging scenarios,so as to break through the physical limitations of the classical imaging systems.The correlation imaging system has brought many attractive features,such as the single-pixel detection,lensless imaging,high-order correlation measurements,and has gradually shown the potential to surpass the classical imaging systems in certain aspects,for instance,the capability of imaging at the extremely low-light level,high-resolution and high-dynamic-range image acquisition,etc.So far,the correlation imaging has certainly stepped into the stage of practical development.However,there still exists some technical problems and challenging scenarios in the practical application of correlation imaging,which also reveals the absenses and shortcomings of the current researches in the correlation imaging community.The common problems in applications,including the extremely weak echo signal,time-consuming acquisition,and lack of the capability of tracking and imaging the moving objct.These problems above not only limit the reconstructed image quality,but also restrict the broad application of correlation imaging in various scenarios,and therefore become the urgent problems in the correlation imaging community,which require to be thoroughly investigated and properly solved.This thesis focuses mainly on a series of key problems which would results in the undesirable image reconstructions in practical applications of correlation imaging systems.Based on the theoretical analysis of the origins and physics of the problems,we propose and validate the corresponding solutions to improve the reconstructed image quality properly and effectively.At the same time,our theoretical analysis would also establish a solid theoretical foundation for the experimental demonstration of correlation imaging in many challenging scenarios.The main contents include:Firstly,the theoretical model of correlation imaging with binary sampling is established,and the corresponding image quality improvement methods are proposed.Intuitively speaking,the binary sampling procedure will lead to the loss of information,which restricts the dynamic range and visual quality of reconstructed image.In order to accurately measure the performance of correlation imaging with binary sampling,we develop a image quality assessment model of correlation imaging based on the well-established concepts and methods in parametric estimation theory.The model regards the procedure of correlation imaging as a process of estimating the parameters of source field from the repeated double-armed observations.Based on the construction of a double-armed likelihood function,we apply the Fisher information and the Cramer-Rao lower bound to evaluate the accuracy limit of the parametric estimation model.Furthermore,the model is validated by experiments,and is capable of predicting the optimal threshold of the binary sampling correlation imaging.In addition,in order to compensate for the loss of information caused by the binary sampling,we introduce the dither techniques to improve the image quality of correlation imaging under the presence of extreme loss of information,also discuss the experimental results and the mechanisms in detail.Secondly,the correlation imaging with the sparse spatial frequencies is investigated,and the corresponding image quality improvement methods are proposed.The correlation imaging system with the insufficient sparse spatial frequencies suffers signal overlapping in the spaital domain,which would appear periodic repetitive structures in the reconstructed images.To address this problem,we present a brief review of the approximation of Fraunhofer diffraction,and introduce an estimation method to extract the point-spread-function of correlation imaging systems,and propose a method to effectively improve the image quality of correlation imaging with sparse spatial frequencies.The principle of the method and the experimental results are discussed in detail,and further work on optimising the quality and efficiency of the proposed method is also presented.Last but not least,the correlation imaging of the moving targets is further explored,and two different methods on imaging the moving object is proposed.During the acquisition process,there inevitably exists relative motion between the object and the detectors,which would induce motion blur in the reconstructed correlation images,resulting in the degradation of image quality.To tackle this problem,we theoretically analyze the impact of relative motion on the correlation images,also provide a brief overview of the research progress in correlation imaging community dealing with the motion blur.Based on the hypothesis,we propose a compressive imaging method using the low-order moments to imaging the moving object,and further achieve a technical breakthrough using the point-like detection to tracking and imaging the moving object.Brief introductions to the principles,experimental demonstration and performance analysis of the above two different methods is also presented.