Spatial-temporal binarization encoding basis modulation Fourier single-pixel imaging

To overcome the limitations of slow switching rates in Fourier basis patterns and the challenges associated with maintaining binary encoding of sinusoidal features in Fourier single-pixel imaging (FSI), this paper presents what we believeto be a novel FSI method based on spatial-temporal binary encoding. Our approach employs an innovative image extension algorithm to preprocess the target scene, effectively decomposing grayscale Fourier basis patterns into two binary basis patterns for projection through dynamic thresholding and error diffusion kernels. By implementing a weighted reconstruction strategy, we efficiently derive single-pixel measurements corresponding to the original grayscale basis patterns, facilitating rapid and high-quality image reconstruction. Experimental results demonstrate substantial improvements in image reconstruction quality compared to existing techniques. For simple scenes, the method achieves remarkable enhancements: an 80% increase in structural similarity index (SSIM), a 43% improvement in peak signal-to-noise ratio (PSNR), and a 62% reduction in root mean square error (RMSE). In complex scenes, the method maintains significant performance gains, with SSIM and PSNR improvements of up to 27.9% and 5.5%, respectively, while effectively minimizing RMSE. Under the same number of projections, this method can achieve reconstruction quality comparable to that of FSI with three-step phase-shifting grayscale projections, while significantly cutting sampling time by up to 28.225s at most and at least 1.895s. The proposed method particularly excels in preserving texture details and enhancing background reconstruction, demonstrating its practical applicability and versatility across diverse target scenes in FSI.