Markov chain Monte Carlo sampling based terahertz holography image denoising.

Area of Science:

• Optics and Photonics
• Image Processing
• Scientific Imaging

Background:

• Terahertz digital holography offers unique imaging capabilities due to terahertz wave transmittance.
• Low image clarity in terahertz holography hinders its widespread adoption.
• Existing denoising methods may not adequately preserve image structures.

Purpose of the Study:

• To develop and evaluate a digital image denoising technique for improving terahertz digital holography images.
• To address the challenge of low image clearness in terahertz holographic imaging.
• To enhance the structural fidelity of denoised holographic images.

Main Methods:

• A digital image denoising technique applied to multiframe superposed images.
• Noise suppression model based on Bayesian least squares estimation.
• Markov chain Monte Carlo (MCMC) sampling for algorithm solution, incorporating a weighted mean filter with Gaussian kernel and nonlinear contrast transform.

Main Results:

• The proposed algorithm effectively suppresses noise in terahertz holographic images.
• The method demonstrates good uniformity, comparable to nonlocal means filtering.
• Superior performance in structure preservation was observed compared to existing methods, validated by structural similarity index measurement and peak signal-to-noise ratio.

Conclusions:

• The developed MCMC-based denoising algorithm significantly improves the quality of terahertz digital holography images.
• The technique effectively balances noise reduction with the preservation of crucial image structures.
• This advancement has the potential to promote the wider application of terahertz digital holography imaging.