Self-supervised image quality assessment for X-ray tomographic images of Li-ion battery

Abstract Image perception plays a fundamental role in the tomography-based approaches for microstructure characterization and has deep impact on all subsequent stages of image processing, such as segmentation 3D analysis. The enhancement perception, however, frequently involves observer-dependence, which reflects user-to-user dispersion uncertainties calculated parameters. This work presents an objective quantitative method, uses convolutional neural networks (CNN) quality assessment X-ray tomographic images. With only dozens annotations, our method allows to evaluate directly precisely Different metrics were employed correlation between predicted scores subjective human annotations. evaluation results demonstrate that can be direct tool guide process order produce reliable results. processing thus evolve into robust observer-independent procedure advance towards development efficient self-supervised approach.