THz Super-Resolution Imaging Based on Complex Fast Deconvolution with Incomplete Boundaries

The method of deconvolution is proved to be an effective method to solve the Low-Resolution (LR) problem of THz imaging. Nevertheless, in the case of unknown boundaries, there are problems of incomplete observations in Terahertz (THz) images. These problems may lead to slow speed for standard deconvolution methods or the occurrence of artifacts for fast deconvolution. In this paper, we proposed a new fast deconvolution algorithm for incomplete boundary THz images with FFT to speed up and iteratively alternate the estimation of the unknown pixels to deal with artifacts. The Alternating Direction Method of Multipliers (ADMM) is used for optimization. The reference evaluation indicators illustrate the efficiency and effectiveness of the proposed method. Super-Resolution (SR) results show that the proposed method has good performance in suppressing artifacts and, the convergence of the presented algorithm is verified. The proposed method is believed to address the LR problems for THz imaging and will be applied in many fields.