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

Area of Science:

• Optics and Photonics
• Image Processing
• Computational Imaging

Background:

• Single-pixel imaging (SPI) is challenged by motion blur and artifacts in dynamic scenes due to long acquisition times.
• Existing methods like compressive sampling and motion compensation offer partial solutions but struggle with complex dynamic scenes.
• Simultaneously achieving high temporal resolution and effective motion artifact suppression in SPI remains a significant hurdle.

Purpose of the Study:

• To develop a novel Single-pixel imaging (SPI) scheme that enhances temporal resolution and suppresses motion artifacts in dynamic scenes.
• To integrate optical flow estimation for motion compensation within a sliding-window Hadamard SPI framework.
• To address the challenge of capturing dynamic scenes with SPI without requiring prior motion or background information.

Main Methods:

• Proposed a sliding-window Hadamard SPI scheme to improve temporal resolution.
• Integrated optical flow estimation to extract motion information from single-pixel data.
• Implemented a motion compensation strategy to correct for motion-induced inconsistencies in measurements.
• Validated the method through simulations and hardware experiments under various sampling rates and window configurations.

Main Results:

• The proposed scheme successfully enhances temporal resolution in dynamic scenes.
• Accurate motion estimation and compensation were achieved using optical flow from single-pixel data.
• Significant improvements in the quality and fidelity of SPI in dynamic scenes were demonstrated.
• The method proved effective across different sampling rates and sliding-window step configurations.

Conclusions:

• The novel sliding-window Hadamard SPI with optical flow compensation effectively tackles motion artifacts in dynamic scenes.
• This approach represents a significant advancement in SPI for dynamic scene imaging, offering high temporal resolution and artifact suppression.
• The method's ability to perform motion compensation without prior information makes it broadly applicable to complex dynamic imaging scenarios.