Optimizing energy dispersive X-Ray Spectroscopy (EDS) image fusion to Scanning Electron Microscopy (SEM) images

Zina-Sabrina Duma; Tuomas Sihvonen; Jouni Havukainen; Ville Reinikainen; Satu-Pia Reinikainen

doi:10.1016/j.micron.2022.103361

Optimizing energy dispersive X-Ray Spectroscopy (EDS) image fusion to Scanning Electron Microscopy (SEM) images

Micron. 2022 Dec:163:103361. doi: 10.1016/j.micron.2022.103361. Epub 2022 Oct 4.

Authors

Zina-Sabrina Duma¹, Tuomas Sihvonen², Jouni Havukainen³, Ville Reinikainen², Satu-Pia Reinikainen²

Affiliations

¹ LUT University, Department of Computational Engineering, Yliopistonkatu 34, Lappeenranta 53850, Finland. Electronic address: Zina-Sabrina.Duma@lut.fi.
² LUT University, Department of Computational Engineering, Yliopistonkatu 34, Lappeenranta 53850, Finland.
³ LUT University, Department of Environmental Engineering, Yliopistonkatu 34, Lappeenranta, 53850, Finland.

PMID: 36219986
DOI: 10.1016/j.micron.2022.103361

Abstract

Fusion and quality enhancement of the low-resolution Energy Dispersive X-ray Spectroscopy (EDS) maps to Scanning Electron Microscopy (SEM) panchromatic images has been proven effective by various pansharpening algorithms. The present paper aims to target the preprocessing of these maps to enhance the efficiency of the pansharpening process, with as little information loss on the chemical distribution, and as little propagated noise as possible. EDS maps present different noise intensities depending on the flatness of the surface of the analyzed object. The uneven surface maps have limited analytical value due to the noise and have not been resolution-enhanced with pansharpening due to the noise propagation limitation. In this paper, different preprocessing methods are evaluated for enabling uneven-surface particles to pansharpening: background removal, upsampling, and noise filtering. The sequence of applying preprocessing steps is analyzed. The optimal order of preprocessing steps is (i) background removal, (ii) noise filtering, and (iii) interpolation. A methodology for each of these steps is presented in the paper. The best performing pansharpening methodology is chosen to be Affinity for individual map analysis and Wavelet for multi-elemental fusion purposes. Following the methodology results in high-resolution EDS maps, even for uneven-surface particles which are, for the first time in literature, subjected to pansharpening.

Keywords: Energy dispersive X-ray spectroscopy; Image fusion; Multivariate image processing; Pansharpening; Scanning electron microscopy.