Construction of simplified boundary surfaces from serial-sectioned metal micrographs

Area of Science:

• Materials Science
• Computer Vision
• Computational Geometry

Background:

• Extracting 3D surfaces from segmented 2D cross-sections is crucial for material analysis.
• Existing methods often produce complex surfaces with artifacts, hindering accurate representation.

Purpose of the Study:

• To present a novel method for accurate 3D boundary surface extraction from segmented cross-section images.
• To address limitations of existing surface extraction techniques, specifically staircase artifacts and excessive triangle counts.

Main Methods:

• Utilized a constrained Potts model for interpolating region boundaries between segmented images, creating a segmented volume.
• Employed marching tetrahedra methods for initial surface extraction.
• Developed a voxel-accurate simplification algorithm to reduce surface complexity and eliminate artifacts.

Main Results:

• Successfully generated triangulated boundary surfaces from segmented volume data.
• Significantly reduced surface complexity by an order of magnitude using the simplification algorithm.
• Applied the method to construct polycrystal grain boundary surfaces from tantalum micrographs.

Conclusions:

• The proposed boundary interpolation and simplification methods offer a novel approach to surface extraction.
• The technique effectively reconstructs complex 3D structures, such as polycrystal grain boundaries, with improved accuracy and reduced complexity.