Fourier single-pixel imaging reconstruction network for unstable illumination

Fourier single-pixel imaging is a computational imaging technique that achieves high-quality imaging of target scenes by measuring the frequency spectrum coefficients of the scene. However, in unstable pulse laser illumination environments, Fourier single-pixel imaging is susceptible to the instability of laser’s power, resulting in a drastic degradation in imaging quality. Additionally, the multiple effects of noise and under-sampling further exacerbate the degradation of the quality of the imaging results. Although multi-pulse accumulation can mitigate these effects to some extent, it significantly increases imaging time, affecting real-time imaging. To address this issue, a Fourier single-pixel imaging reconstruction network for unstable illumination is proposed. The proposed method is mainly composed of a spectrum correction module, a noise estimation module and a denoising module. The spectrum correction module is used to eliminate the effect of unstable pulse lasers on imaging quality, reducing image distortion caused by laser pulses. The noise estimation module is utilized for noise estimation, while the denoising module ultimately eliminates the effect of noise to achieve high-quality reconstruction results. Simulation and experimental results demonstrate that the proposed method not only eliminates the effect of unstable pulse lasers on reconstruction results but also excels in noise processing, showcasing outstanding performance in recovering high-quality Fourier single-pixel imaging results.