High-quality Fourier single-pixel imaging via a block-scanning approach

Abstract

Fourier single-pixel imaging (FSI) has attracted wide attention as a promising spatial imaging technique. However, Full sampling requires a substantial collection of Fourier-basis patterns. An under-sampling strategy leads to losing image details, resulting in poorer reconstruction quality. To this end, this work proposes a novel FSI method called Block-Scan FSI (BSFSI) to address this problem. Specifically, we establish a comprehensive error model involving system and random error. The established model unveils that reducing the frequency range of modulation patterns can improve image quality by suppressing systematic and random errors. Accordingly, we present a block-scanning structured illumination strategy to reduce error levels and construct a flexible BSFSI system for implementing this block-scanning strategy. Simulation and experimental results showed that the proposed BSFSI method uses only 25% of the pattern types while achieving an impressive 14% improvement in image quality compared to the traditional FSI method.