Polarization prior to single-photon counting image denoising

Area of Science:

• Optics and Photonics
• Image Processing
• Computational Imaging

Background:

• Single-photon counting (SPC) imaging excels at detecting targets in extremely low light.
• Traditional denoising methods struggle with low light due to poor signal-noise distinction.
• Existing techniques often rely on image-derived priors, limiting denoising performance.

Purpose of the Study:

• To develop a novel and effective method for single-photon counting (SPC) image denoising.
• To leverage polarization information for enhanced noise reduction in low-light imaging.
• To improve the signal-to-noise ratio (SNR) and spatial detail preservation in SPC images.

Main Methods:

• Proposed a novel polarization prior for SPC image denoising.
• Utilized a special polarization SPC (PSPC) image with a higher SNR.
• Developed an 'anti-noise' dictionary from high-SNR PSPC image patches.
• Formulated a denoising model using sparse representation, Poisson log-likelihood, and total variation constraints.
• Solved the model using an efficient variable splitting method.

Main Results:

• The proposed polarization prior effectively transfers spatial details from PSPC to SPC images.
• Demonstrated improved denoising performance in both simulated and real low-light data.
• Achieved higher SNR compared to traditional denoising approaches.
• Verified effectiveness through visual comparison and quantitative analysis.

Conclusions:

• The novel polarization prior significantly enhances SPC image denoising.
• The method effectively distinguishes signal from noise in extremely low light conditions.
• This approach offers a promising solution for improving image quality in low-light imaging applications.