USTSR: a U-shaped swin transformer framework for super-resolution in solid-state THz array imaging systems.

Area of Science:

• Optics and Photonics
• Artificial Intelligence
• Image Processing

Background:

• Terahertz (THz) imaging offers potential for security, biomedical, and industrial applications.
• Current THz imaging is limited by low resolution, noise, and slow scanning speeds.
• Existing methods struggle with diffraction artifacts and spectral resolution.

Purpose of the Study:

• To develop an integrated THz imaging approach for real-time, high-resolution imaging.
• To enhance image quality by suppressing noise and artifacts without complex hardware.
• To introduce a novel evaluation strategy for THz image quality assessment.

Main Methods:

• Integrated a solid-state frequency multiplier-based focal-plane THz array system for real-time imaging.
• Developed a transformer-based super-resolution neural network (USTSR) specifically for THz image enhancement.
• Implemented a dual-metric evaluation strategy using MS-SSIM and FID for THz A-scan quality.

Main Results:

• USTSR significantly enhances image resolution and suppresses noise tailored for THz imaging.
• The integrated system enables rapid, real-time THz imaging.
• Achieved superior performance with 33.17 dB PSNR and 0.86 SSIM compared to state-of-the-art methods.

Conclusions:

• The developed integrated THz imaging system and USTSR offer robust, high-fidelity imaging capabilities.
• This approach overcomes limitations of conventional THz imaging, enabling faster and more accurate measurements.
• Demonstrates broad applicability for practical measurement tasks in security, diagnostics, and testing.