Imaging involves the process of recording and reproducing images as close to reality as possible, encompassing both direct and indirect approaches. In direct imaging, the object is directly recorded. Coded aperture imaging (CAI) is an example of indirect imaging, that utilizes optical recording and computational reconstruction to obtain information about an object. Computational reconstruction can be achieved using different linear, nonlinear, iterative, and deep learning algorithms. In this study, we proposed and demonstrated two computational reconstruction algorithms based on the non-linear Lucy-Richardson algorithm (NL-LRA), one for limited support images and another for full-view images based on entropy reduction. The efficacy of these algorithms has been validated through simulations and optical experiments carried out in visible and infrared (IR) light with different coded phase masks. The methods were also tested on a commercial IR microscope with internal GlobarTM and synchrotron sources. The results obtained from the two algorithms were compared with those from their parent methods, and a notable improvement in both entropy and the convergence rate was observed. We believe the developed algorithms will drastically improve image reconstruction in incoherent imaging applications.
原文链接
分享
引用
