Scalable Double Regularization for 3D Nano-CT Reconstruction.

Area of Science:

• Geoscience
• Imaging Science
• Computational Science

Background:

• Nano-computed tomography (Nano-CT) is vital for studying sub-micron pore structures in shale, essential for oil and gas evaluation.
• Nano-CT reconstruction faces challenges like low signal-to-noise ratio, geometric misalignment, and large datasets.
• Existing methods often reconstruct data slice-by-slice, limiting comprehensive analysis.

Purpose of the Study:

• To develop a novel, scalable method for 3D structural reconstruction from Nano-CT data.
• To address the unique challenges posed by Nano-CT imaging of shale.
• To improve the accuracy and efficiency of pore structure analysis in shale formations.

Main Methods:

• Proposed a scalable double regularization (SDR) method utilizing total variation and L1 regularization.
• Implemented simultaneous 3D reconstruction across slices, enabling information borrowing.
• Developed a memory-efficient algorithm exploiting data sparsity and consistent geometry.

Main Results:

• The SDR method effectively reconstructs 3D structures from Nano-CT data.
• Demonstrated superior performance over alternative methods on synthetic and real shale datasets (JLD and LMX).
• Achieved significant visual and quantitative improvements in reconstruction accuracy.

Conclusions:

• SDR offers a robust and scalable solution for Nano-CT 3D reconstruction.
• The method enhances the understanding of shale pore structure for energy resource development.
• This approach advances non-destructive imaging techniques in geoscience applications.