Accelerated gradient restarted FISTA-TV algorithm for 3D terahertz transmission tomography

Area of Science:

• Physics
• Engineering
• Materials Science

Background:

• Terahertz computed tomography (THz-CT) is a valuable non-destructive technique for analyzing complex sample structures.
• Existing THz-CT methods face challenges in reconstruction quality, especially with limited projection data.

Purpose of the Study:

• To develop an advanced THz-CT system and a novel algorithm for enhanced imaging.
• To improve the quality of THz-CT reconstructions under sparse projection conditions.

Main Methods:

• A 2.5 THz continuous-wave THz-CT system was built, incorporating a quantum cascade laser and a high-speed focal plane array detector.
• A fast iterative shrinkage-thresholding algorithm with total variation regularization (FISTA-TV) was proposed and implemented.

Main Results:

• FISTA-TV significantly outperformed conventional methods in suppressing noise and streak artifacts.
• The algorithm effectively preserved edge sharpness and internal structural details.
• Quantitative metrics (Laplacian variance, mean gradient magnitude, entropy) confirmed enhanced sharpness and reduced noise.

Conclusions:

• FISTA-TV is a feasible algorithm for high-quality THz-CT imaging, particularly under limited-angle conditions.
• The developed THz-CT system and FISTA-TV show potential for non-destructive inspection of polymers and pharmaceuticals.