Practical Research Of Ghost Imaging Configurations

Ghost imaging(GI), which is di?erent from the conventional imaging tech-nology, utilizes second-order correlation of light ?eld. In some con?gurations ofGI, just a bucket detector, which has no spatial resolution, is needed to retrievethe image of the object. GI shows superiorities to the traditional imaging due toits features such as free from the atmospheric disturbance and lensless imaging.These advantages make GI be important and practical in real applications ofwide range of areas. In GI, the image comes from computation. We analysis anddemonstrate GI in the computational view, and give the improvements to GI inorder to promote GI to the practical applications.The main work of this article is the research about some con?gurations ofGI in real applications. The main results are as follows:(1) We have analyzed the in?uence of noise in GI, and presented a theo-retical analysis together with an experimental demonstration of a new techniquewhich we call iterative denoising of ghost imaging. An appropriate threshold isapplied in the iteration to eliminate the detrimental in?uence of coherence. Inthe iterative denoising of ghost imaging, no adjustments of the GI setup andprior knowledge of objects are needed. Our results indicate that IDGI gives amuch better performance than DGI even with a few number of iterations.(2) We analysis the relation between the number of measurements and thesignal to noise ratio(SNR) of GI. When the coherence length of the light isequal to the size of a pixel, the SNR will increase linearly with measurements,however, when the coherence length of the light is larger than the size of apixel, the SNR advances gradually to an asymptotic line with the increasingof the measurement number. In practice, when the coherence length is large,the number of measurements should be restricted to an appropriate value sinceenormous number of measurements is meaningless.(3) In the context of a new con?guration of GI, which is called correspon-dence imaging, we utilize the correlation coe?cients to separate the referencedetector speckle patterns into positive and negative correlated parts. A positiveas well as a negative image of the object are obtained by averaging over corre-sponding the speckle patterns. The visibility and contrast-to-noise ratio of thepositive image are discussed, and it is found that the latter will reach a maximumby averaging over a little less than half of the total number of reference specklepatterns.(4) We demonstrate sub-Rayleigh limit imaging of an object via intensitycorrelation measurements. The resolution of this sub-Rayleigh limit imaging sys-tem is only tied to the lateral resolution of the illumination light. The imagecompletely una?ected by the disturbance of di?raction-limit is achieved underthe condition that the imaging system has an appropriate ?eld of view. Variousreconstruction algorithms proposed for intensity measurements can be utilized.We demonstrate that, by undersampling scanning object with a reconstructionalgorithm related to compressed sensing, an image with the resolution exceedingthe ?nest resolution de?ned by the NA of the system can be obtained. Ex-perimental results show that the measurements needed to achieve sub-Rayleighresolution enhancement can be less than 10% of the pixels of the object.