Improving reconstructions in nanotomography for homogeneous materials via mathematical optimization

Area of Science:

• Materials Science
• Image Processing
• Computational Physics

Background:

• Compressed sensing reconstructs high-quality images from limited data using prior knowledge.
• Nanotomography requires advanced reconstruction techniques for nanoscale imaging.
• Existing methods may struggle with limited projection data in nanotomography.

Purpose of the Study:

• To enhance compressed sensing models for nanotomography image reconstruction.
• To introduce novel algebraic inequalities as additional constraints.
• To improve reconstruction quality and efficiency, particularly with sparse data.

Main Methods:

• Developed new classes of algebraic inequalities for compressed sensing.
• Incorporated pixel brightness upper bounds and density variation penalties.
• Utilized interior point methods for solving the optimization models.
• Validated models on simulated and experimental electron tomography and nano-CT data.

Main Results:

• The enhanced models were optimized quickly and demonstrated improved reconstruction quality.
• Outperformed existing image reconstruction methods on both simulated and experimental datasets.
• Showed superior performance particularly when using a limited number of tilt angles.
• Experimental validation included macroporous zeolite and copper microlattice structures.

Conclusions:

• The proposed enhanced compressed sensing models significantly improve nanotomography image reconstruction.
• The inclusion of problem-specific knowledge, like brightness bounds and density constraints, is effective.
• This approach offers a promising solution for high-quality imaging with sparse projection data in nanotomography.