Three-dimensional reconstruction using variable exponential function regularization for wide-field polarization...

Area of Science:

• Optical Physics and Imaging
• Computational Science and Engineering

Background:

• Diffuse polarization methods enable 3D surface reconstruction using diffuse reflection light polarization.
• Nonconvex particle surfaces and scanning electron microscope (SEM) stripe noise limit reconstruction accuracy, causing local optima and distorted textures.

Purpose of the Study:

• To develop an adaptive 3D reconstruction method addressing nonconvexity and inclination of particle surfaces.
• To mitigate gradient unintegrability issues arising from specimen skew and undulation.

Main Methods:

• Proposed a variable exponential function regularization method to fit particle surface functions.
• Developed an adaptive 3D reconstruction strategy based on this regularization technique.
• Validated the method using finite-difference time-domain (FDTD) simulations and experimental imaging.

Main Results:

• The variable exponential function regularization effectively handles nonconvex surfaces and inclination.
• The adaptive method overcomes local optimal solutions common in particle surface reconstruction.
• Stripe noise from SEM depth images is significantly reduced, preserving surface texture.

Conclusions:

• The proposed adaptive 3D reconstruction method enhances accuracy for complex particle surfaces.
• Variable exponential function regularization provides a robust solution for challenging surface topography reconstruction.
• The technique demonstrates effectiveness in both simulated and real-world imaging scenarios.