Zeolites are an important group of materials with a wide range of application in the catalysis industry. Many structural studies of zeolites rely on high resolution electron microscope imaging . However, due to their high sensitivity to electron irradiation, zeolites deteriorate quickly under exposure to the electron beam. Low-dose imaging techniques use a reduced electron flux to slow the crystal degradation process, which gives more time for adjustment of the microscope configuration and better control over the progression of damage. However the disadvantage of low-dose imaging is poor signal to noise ratio which is often alleviated by averaging multiple image frames in a time series for improved image quality. Traditional rigid cross-correlation function (XCF) image registration methods work well for aligning high-dose time series of radiation-robust materials which experience little or no deformation during image acquisition. However, the deformation in radiation-sensitive materials, often manifest by sample shrinkage, means that the single translational shift vector from rigid image registration may not be sufficient for aligning time series and hence a non-rigid registration scheme is needed.
In this work, a low-dose time series of ZSM-5 zeolite consisting 60 image frames were recorded using an aberration-corrected JEOL2200MCO TEM (Figure 1). Two registration methods, a rigid XCF registration and a new non-rigid registration, were used to align the series respectively. The non-rigid registration method  is assisted by an IQ factor criterion, which evaluates the quality of the averaged image of the series as the registration proceeds and selects the best averaged image as the reference for future registration iterations.
The results show that the new non-rigid registration is helpful for alignment of low-dose TEM image series of radiation-sensitive materials that experience deformation during imaging, especially when the number of frames is small and when the sample is already damaged (Figure 2). This implies that, for TEM image series, the non-rigid registration approach is more effective in noise suppression and in avoiding the image components of a damaged sample compromising the final averaged image.
To cite this abstract:Chen Huang, Benjamin Berkels, Angus Kirkland; Non-rigid image registration of low-dose image series of zeolite materials. The 16th European Microscopy Congress, Lyon, France. https://emc-proceedings.com/abstract/non-rigid-image-registration-of-low-dose-image-series-of-zeolite-materials/. Accessed: July 5, 2020
EMC Abstracts - https://emc-proceedings.com/abstract/non-rigid-image-registration-of-low-dose-image-series-of-zeolite-materials/