Motion compensation for dynamic single-pixel imaging via optical flow in sliding windows

Abstract

Single-pixel imaging (SPI) faces a primary challenge in dynamic scenes due to its inherently long acquisition time, which readily induces motion blur and artifacts. Compressive sampling approaches can enhance frame rates, while motion compensation approaches can reduce motion blur. However, developing an SPI scheme that simultaneously achieves high temporal resolution and effective motion artifact suppression for complex dynamic scenes remains a challenge. We propose a scheme based on sliding-window Hadamard SPI, integrating optical flow estimation for motion compensation. This scheme firstly enhances temporal resolution via sliding-window sampling, then accurately estimates scene motion using optical flow information extracted from the single-pixel data, and subsequently compensates for the motion-induced inconsistencies in single-pixel measurements. To the best of our knowledge, this is the first work to apply a motion compensation strategy that does not require prior motion and background information to SPI of complex dynamic scenes. Simulation analyses and practical hardware experimental results collectively validate the effectiveness of the proposed method under various sampling rates and sliding-window step configurations, demonstrating its capability to significantly enhance the quality and fidelity of SPI in dynamic scenes.