Understanding Deformation Motion of Colloidal Nanosheets from CLSM Images using Deep Learning-based Approach

This paper considers a problem of understanding deformation motion of colloidal nanosheets from a set of confocal laser scanning microscopy (CLSM) images corrupted by noises. First, we present a robust method for detecting nanosheet objects from noisy CLSM images by introducing the deep learning-based approach. Then, we develop a method for understanding motions of nanosheet objects in colloid liquid. Such a method is constituted by introducing the idea of the so-called gradient-based feature descriptor, in which the local and global deformation motions are effectively visualized. The performance is demonstrated by some experimental studies.