Computational evaluation of sparse coding on off-axis electron holograms: comparison between charge-coupled device...

Area of Science:

• Electron Microscopy
• Computational Imaging
• Materials Science

Background:

• Off-axis electron holography is crucial for characterizing materials at the nanoscale.
• Image quality in electron holography is often limited by noise and low electron doses.
• Sparse coding offers a potential solution for noise reduction and signal enhancement.

Purpose of the Study:

• To computationally evaluate the effectiveness of sparse coding for image inpainting and denoising in electron holography.
• To compare the performance of charge-coupled device (CCD) and direct-detection device (DDD) cameras under varying noise conditions.
• To assess the impact of sparse coding on hologram quality at low electron doses.

Main Methods:

• Simulated off-axis electron holograms of a phase step object (semiconductor p-n junction).
• Modeling of noise characteristics for CCD (shot, dark-current, read-out) and DDD (shot noise only) cameras.
• Application of sparse coding techniques to simulated holograms across a range of electron doses.
• Analysis of interference fringe visibility in both original and sparse-coded holograms.

Main Results:

• Sparse coding successfully recovered meaningful interference fringes even from holograms with imperceptible fringes.
• Simulated CCD and DDD holograms showed visible fringes only at higher electron doses (1 and 0.01, respectively).
• The combination of DDD cameras and sparse coding reduced the necessary electron dose by over a thousandfold compared to CCD cameras without post-processing.

Conclusions:

• Sparse coding is highly effective for denoising and inpainting electron holograms, significantly improving image quality.
• The integration of direct-detection device cameras with sparse coding enables electron holography at substantially lower electron doses.
• This approach promises advancements in lower-dose and higher-speed electron holography for advanced materials characterization.