Mitigation of Systematic Errors in EFTEM Spectral Imaging via Median Filtering

A method has been developed for the removal from CCD images of point blemishes (PBs) ubiquitous in the acquisition of energy-filtered transmission electron microscopy (EFTEM) images. PBs arise from spurious radiation, such as X rays, impinging on the CCD detector during imaging acquisition, and form a major source of systematic error in the multivariate statistical analysis of EFTEM spectral images (SI). The method involves median filtering among images acquired sequentially, and concomitant oversampling in the energy coordinate, to remove these point blemishes without removing real spectroscopic image contrast. The method removes a major obstacle to automated nanometer-resolution chemical imaging of individual nanostructures for the purposes of statistical sampling or tomography.

Intended impact: Automated nm-resolution chemical imaging of individual nanostructures for statistical sampling or tomography

Objectives: Mitigation of significant source of systematic error in the acquisition and analysis of EFTEM spectral images

Goals:

• Robust removal of point blemishes ubiquitous in acquisition of CCD images in the TEM
• Noise filtering of low-signal CCD images without loss of spatial resolution

Technical approach:

• Median filtering of sequentially acquired images with redundant or largely redundant image contrast
• Oversampling of acquired EFTEM spectral images, consistent with number of median filtered images, to capture relevant spectral response
• Acquisition of unprocessed CCD images for EFTEM spectral imaging, and application of dark current and gain normalization procedures following median filtering

• Robust removal of point blemishes in CCD images arising from spurious high-energy electromagnetic radiation
• Reduction in noise structure of spectral images to below Poisson limit
• Removal of significant barrier to automation of chemical imaging and tomography
• MatLab scripts that apply the method