Impact of uniform beam illumination on compressed sensing-based terahertz single pixel imaging system

Abstract Compressive sensing has significantly advanced terahertz imaging by enabling high-quality image reconstruction from a reduced number of measurements, paving the way for faster and more efficient acquisition. However, the quality of the reconstructed images remains highly sensitive to experimental conditions, particularly the spatial non uniformity of the illuminating beam. Non uniform beam distribution can cause variations in the spatial intensity across the imaged area, leading to inconsistent sampling and loss of crucial information during the reconstruction process. In this work, we present a novel, cost-effective beam shaping strategy using a 3D-printed 2D Powell collimating lens to produce a uniform THz beam profile across the mask area. Unlike conventional Gaussian-beam-based setups that require additional normalization and correction steps, our approach directly improves the spatial sampling consistency without added system complexity. Experimental results demonstrate that this physical beam homogenization significantly enhances image reconstruction quality, both quantitatively and perceptually. The proposed solution thus offers a practical and scalable enhancement to single-pixel THz compressive imaging systems, with strong potential for real-time and field-deployable applications.