Explicit Compression Degradation Estimations for Low-Sampling Single-Pixel Imaging using Hadamard Basis.

Area of Science:

• Optics and Photonics
• Computational Imaging
• Signal Processing

Background:

• Single-pixel imaging (SPI) acquires 2D images from 1D measurements, but quality is limited by the number of samples.
• Compressive sensing (CS) enables reconstruction from undersampled SPI data, yet explicit degradation models remain unclear.
• Existing restoration methods rely on implicit priors or data-driven approaches, lacking explicit compression modeling.

Purpose of the Study:

• To present a degradation estimation technique for explicit compressive sampling modeling in low-sampling SPI.
• To improve SPI reconstruction quality by understanding and accounting for compression-induced degradations.
• To demonstrate the applicability of the proposed method for dynamic scenes and single-pixel video imaging.

Main Methods:

• Developed a degradation estimation technique to explicitly describe compressive sampling in SPI using Hadamard basis patterns.
• Proposed a self-supervised learning method to estimate explicit degradation models, primarily blur kernels.
• Investigated the impact of varying sampling ratios on compression degradation models and SPI results.

Main Results:

• Compression degradation models were successfully characterized and reflected at different sampling ratios.
• Explicit blur kernels, varying with sampling ratios, were estimated using self-supervised learning.
• Numerical and experimental demonstrations confirmed the effectiveness of the approach for SPI reconstruction and video imaging.

Conclusions:

• The proposed degradation estimation technique provides an explicit model for compressive sampling in SPI.
• This method significantly enhances low-sampling SPI reconstruction by accounting for blur kernels.
• The approach shows promise for advancing practical applications of SPI, including dynamic scene imaging.