Constructing High-Resolution Microstructure Image for Cement Hydration Using Real-ESRGAN

Observing the microstructure images of cement can help us understand cement hydration and assist in designing such materials. Through the use of microscopic imaging equipment, such as scanning electron microscopy, images of cement microstructure can be captured. However, in the actual observation process, due to the limitations of the imaging equipment, it is common to adjust the equipment focus and observe different positions, but only high-resolution images of partial area can be obtained. It will increase economic costs and manual repeat scanning costs for obtaining whole sample vision. This paper studies the construction of high-resolution microstructure image for cement hydration based on Real-ESRGAN. This method utilizes the adaptability, parallelism and fault-tolerance of intelligent computing, which provide us a potential and low-cost way to obtain high-resolution microstructure image compared to microscopic imaging equipment. Experiment results demonstrate that the proposed method presents a good performance in constructing high-resolution microstructure image of cement hydration.