KAFE-SPN: attention fusion enhanced-based single-pixel phase-shifting holography

This paper proposes a single-pixel phase-shifting holographic imaging method based on the Kolmogorov-Arnold attention fusion enhanced mechanism to overcome traditional digital holography's limitations in single-pixel imaging. A Kolmogorov-Arnold attention fusion-enhanced based single-pixel phase-shifting network (KAFE-SPN) is proposed, establishing a 1D compressed data-to-multi-phase-shifted hologram mapping framework for the generation of four holograms with 0, π/2, π, and 3π/2 phase shifts. This method does not rely on a mechanical phase shifter or multiple data acquisitions, significantly reducing the amount of imaging data while retaining high resolution and depth information. Experimental results demonstrate that KAFE-SPN generates high-quality holograms at 20% sampling rate and maintains stable performance under extremely low sampling rates and high noise levels. It is expected to provide a high-precision 3D reconstruction approach for related holographic imaging fields.