Terahertz Image Super-Resolution Reconstruction Based on Complex Deconvolution Algorithm with Different Criteria

Super-Resolution (SR) is one of the most fundamental issues in THz imaging processing. Deconvolution is widely used in THz imaging SR problems. Heavy-tailed distribution has good performance in deconvolution for its capability to preserve edges. By making use of the wave nature of THz waves, complex deconvolution is employed for solving THz images. A complex deconvolution algorithm based on criteria is proposed in this paper. And three criteria are discussed to get the most optimal result. Compared with the traditional methods, which are set the index as a fixed value, our method is more flexible for obtaining clear edges. The measured PSF is used to analyze the THz image by a complex image blurred model. Experiments illustrate the efficiency and accuracy of the proposed method. All three criteria can solve the LR problem well. Relative to Peak Signal to Noise Ratio (PSNR) and information entropy, the SR result of Structural Similarity Index Measure (SSIM) is better visual. Thus, we prefer SSIM as the criterion in our work