Quantitative THz time-domain single-pixel imaging using statistical scale-adaptive computational ghost imaging with random masks

We present a terahertz time-domain spectroscopy (THz-TDS) approach integrated with single-pixel imaging (SPI) using a metallic random mask, enabling simultaneous acquisition of spatial and spectral information with high quantitative accuracy. While computational ghost imaging (CGI) provides excellent noise robustness, it typically yields relative correlation values, losing the quantitative scale proportional to the THz electric field. Conversely, ordinary least squares (OLS) reconstruction maintains quantitative scaling but is highly susceptible to noise due to the ill-conditioned nature of random masks. To overcome this trade-off, we introduce a statistical scale-adaptive computational ghost imaging (SSA-CGI) method that hybridizes the correlation-based stability of CGI with the quantitative scaling of OLS. This approach statistically transfers the quantitative voltage scale from OLS to the high-SNR CGI reconstruction, enabling accurate amplitude and phase mapping. The proposed technique was experimentally validated using a metallic aperture and patterned SU-8 structures, demonstrating spatially resolved spectroscopic imaging with high scaling fidelity and yielding group refractive indices consistent with reported values. Owing to its simple and robust configuration, the SSA-CGI method offers a practical, low-cost route toward quantitative THz spectroscopic imaging for applications in material characterization, nondestructive testing, and biomedical diagnostics.