Abstract: Fourier single-pixel imaging technology is one sort of single-pixel imaging technology that provides both outstanding image quality and high imaging efficiency. When utilizing this technology, an efficient sampling approach is crucial to enhance the reconstruction quality for images with various attributes. In this work, the implementation principles and processes of three commonly used sampling methods for Fourier single-pixel imaging, namely circular sampling, Gaussian random sampling, and adaptive sampling, are studied. Taking optimal sampling as a reference, the quality of reconstructed images of the four sampling methods at various sample rate is simulated and compared, with a focus on the benefits and drawbacks of the four sampling techniques in processing images with various attributes. The results demonstrate that for images with various spectral energy distribution characteristics, the four sampling methods exhibit distinct peak signal-to-noise ratios in reconstructed images, and each sampling method has its own optimal applicability. This result will provide important reference for how to efficiently select the best sampling method to reconstruct the images with various attributes.

Key words: image and information processing, Fourier single-pixel imaging, circular sampling, Gaussian random sampling, adaptive sampling