Frequency-dependent PI parameter optimization for enhanced terahertz near-field imaging quality

Abstract

The trade-off between scanning speed and image quality constrains the application of terahertz near-field imaging in high-throughput scenarios. In this study, quantitative models were established to link the proportional (P) and integral (I) parameters to image quality—quantified by the Brisque score—at 0.4 Hz and 0.5 Hz using second-order polynomials (R 2 = 0.65 and 0.60). The optimal parameters were found to be frequency-dependent. At a scanning frequency of 0.4 Hz, the optimal proportional gain P and integral gain I were 560 and 575, respectively, yielding a minimum Brisque score of 35.70. At 0.5 Hz, the optimal values shifted to P  = 100 and I = 225, with a corresponding minimum score of 41.11. With these optimized parameters at 0.5 Hz, the Brisque score decreased from 49.10 (obtained with conventional settings, P  = 400, I = 300) to 41.11, representing a 16.3% relative improvement, while the scanning speed increased by 25%. This work offers a quantitative framework for PI parameter optimization in terahertz near-field imaging, enabling fast, high-quality imaging in biomedical and other applications. The frequency-dependent trends reported are based on experiments conducted at 0.4 Hz and 0.5 Hz, with supplementary validation at 0.3 Hz and 0.6 Hz.