Performance Evaluation of Spatial Modulation Patterns in Compressive Sensing Terahertz Imaging

Terahertz (THz) imaging traditionally uses pixel-to-pixel mechanical raster scanning, which is slow and limits the resolution and potential applications. Compressive Sensing Terahertz (CS-THz) imaging has the potential to solve these challenges by reducing the number of required measurements. However, there is an existing research gap between the current CS-THz implementations, which often use random binary masks without further consideration, and specialized Compressive Sensing works, which focus on RIP and coherence reduction but ignore the physics underlying THz wave propagation such as the effect of diffraction on masks patterns. This paper discusses and evaluates the use of low-coherence sensing matrices as mask patterns for CS-THz imaging. Although not previously used in THz imaging, Best Antipodal Spherical Code masks show the best image reconstruction performance among the considered alternatives. We also demonstrate the feasibility of phase-only modulation.