Multi-Frequency Resolution-Oriented Reconstruction for Single-Pixel Terahertz Imaging

Single-pixel terahertz (THz) imaging enables spatial reconstruction using a single detector and a sequence of coded modulations. However, practical systems often exhibit non-ideal sensing operators due to amplitude-only modulation, diffraction, and frequency-dependent propagation. This paper presents a practical multi-frequency reconstruction framework for experimentally calibrated single-pixel THz sensing operators, implemented in a matrix-free form that avoids explicit stacking across frequencies. By using a resolution-oriented penalized inversion solved via reweighted quadratic updates, the results show consistent improvements in reconstruction fidelity and target contrast over matched filtering, total variation-regularized and least-squares baselines as the measurement count and number of frequency samples increase.