Plasma Focused Ion Beam Curtaining Artifact Correction by Fourier-Based Linear Opti-mization Model

Abstract

Focused ion beam scanning electron microscope tomography is a destructive slice-and-image technique for obtaining three-dimensional structural information. Xenon plasma focused ion beam (PFIB) is a promising new ion source technology that has a much higher material removal rate than traditional gallium source technology -greatly increasing the analyzable volume of material [1].Unfortunately, due to the heterogeneous nature of any interesting sample, plasma milling rates vary, causing vertical ripples ("curtains") on the imaged cut face of the sample. These curtains appear as dark or light straight-line artifacts on secondary and backscatter electron images. In order to perform an accurate automated quantitative analysis of the resulting tomogram, these artifacts must be reduced as much as possible through physical means, such as using a deposited capping layer on the top of the sample or through a "rocking polish" technique which works by milling the sample at two discrete angles between image acquisitions. While these techniques help to reduce the ef-fect, curtains remain at two angles and need to be further reduced through image processing means.We have developed a new algorithm to effectively correct these image artifacts by applying a Fourier-based linear optimization model to detect and remove curtains at two discretized angles. The algorithm and linear optimization model was implemented in the Anaconda Python 3.5.2 distribution using the open-source Gnu Linear Programming Kit (GLPK) to solve the linear optimization problem on small image blocks, then stitch the corrected image back together.In order to test the algorithm, we ran the algorithm on a ultra-high performance concrete tomogram taken on an FEI PFIB. A backscatter electron detector was used to take images at 48nm × 48nm pixel resolution with a 50nm slice thickness. During imaging, a rocking mill was used, leaving curtains at approximately 1 • and -7 • from the vertical. We have shown that our algorithm is able to effectively detect and correct curtaining artifacts without causing changes to other structures (see figures 1 & 2).Future work includes taking advantage of the parallel nature of the algorithm and using faster implementations and solvers in order to speed up the processing. Additionally, by combining multispectral data, such as information from both backscatter and secondary electron images, we believe we can obtain more effective curtain detection and correction. Other works meant to reduce only vertical curtains (such as [2]) take advantage of three-dimensional inter-slice information, which may prove advantageous to our model and should be further explored [3].

Advanced characterisation of 3D structure and porosity of ordinary portland cement (OPC) mortar using plasma focused ion beam tomography and X‐ray computed tomography

“…This is in part due to the variation of their hardness and thus differing milling/sputtering rates 36,37 . The most common image artefact is curtaining that makes following data analysis quite difficult 38 . The success of pore and structure characterisation depends strongly on sample preparation 23,39 .…”

Removing Stripes, Scratches, and Curtaining with Nonrecoverable Compressed Sensing

“…A third approach treats the destriping issue as an ill-posed inverse problem. Prior knowledge is used to regularize an optimization problem (Bouali & Ladjal, 2011) and separate the unidirectional stripes from the image (Liu et al, 2013; Fitschen et al, 2017; Schankula et al, 2018). A similar class of research, known as compressed sensing (CS), has become highly successful toward solving inverse problems with incomplete data by finding maximally sparse solutions—but has yet been applied to remove scratch and stripe artifacts.…”

Removing Stripes, Scratches, and Curtaining with Non-Recoverable Compressed Sensing

“…Prior knowledge is used to regularize an optimization problem (Bouali & Ladjal, 2011) and separate the unidirectional stripes from the image (Liu et al, 2013;Fitschen et al, 2017;Schankula et al, 2018). A similar class of research, known as compressed sensing (CS), has become highly successful toward solving inverse problems with incomplete data by finding maximally sparse solutions -but has yet been applied to remove scratch and stripe artifacts.…”