Terahertz image enhancement method combining contrast limited adaptive histogram equalization and edge information

Area of Science:

• Physics
• Engineering
• Materials Science

Background:

• Terahertz (THz) nondestructive testing (NDT) suffers from low resolution, poor signal-to-noise ratio, and insufficient contrast.
• These limitations hinder the effective analysis and interpretation of THz images in various applications.
• There is a need for advanced image processing techniques to overcome these challenges.

Purpose of the Study:

• To propose and evaluate a novel THz image enhancement method.
• To improve the imaging resolution, contrast, and signal-to-noise ratio of THz images.
• To provide a software-based solution for enhancing THz images for NDT.

Main Methods:

• A hybrid approach combining Contrast Limited Adaptive Histogram Equalization (CLAHE) with edge information extraction.
• Gaussian low-pass filtering for initial noise suppression.
• Canny edge detection for feature extraction.
• Weighted fusion of the enhanced image and edge map for optimal results.

Main Results:

• Significant improvements in objective image quality metrics.
• Mean Gradient (MG) increased by over 66%, indicating enhanced edge sharpness.
• Tenengrad sharpness metric increased by more than 46%.
• Structural Similarity Index (SSIM) maintained at approximately 0.8, preserving image integrity.

Conclusions:

• The proposed method effectively enhances edge sharpness and contrast in THz images.
• It successfully balances detail enhancement with noise suppression.
• This technique offers a valuable software-based solution for improving THz NDT.