Hyperspectral image denoising using the robust low-rank tensor recovery

Area of Science:

• Remote Sensing
• Image Processing
• Data Science

Background:

• Hyperspectral image (HSI) denoising is crucial for accurate analysis.
• Traditional methods struggle with outliers and non-Gaussian noise.
• Existing techniques may compromise the global structure of HSI data.

Purpose of the Study:

• To develop a novel HSI denoising model.
• To address limitations of traditional denoising methods.
• To preserve the global structure of HSI while removing diverse noise types.

Main Methods:

• Utilizing the low-rank tensor property of clean HSI data.
• Leveraging the sparsity of outliers and non-Gaussian noise.
• Employing a robust low-rank tensor recovery model.
• Solving the model using the inexact augmented Lagrangian method.

Main Results:

• The proposed model effectively removes Gaussian noise, impulse noise, and dead lines.
• It successfully preserves the global structural information of HSI.
• Experimental results on simulated and real HSI data confirm its efficiency.

Conclusions:

• The robust low-rank tensor recovery model offers superior performance for HSI denoising.
• This method provides a significant advancement in preprocessing HSI data.
• The approach is efficient and preserves essential HSI characteristics.