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

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