Particle Detection in Nanomaterial Images Based on Normalized Graph Cuts and Binary Segmentation

A method for detecting particles in nanomaterial images based on normalized graph sections and binary segmentation is presented in the article. The method includes the following calculations: preliminary segmentation of the image into regions using the simple linear iterative clustering method; segmentation of the obtained regions by the method of normalized cuts on graphs, binary segmentation of the image based on the automatic selection of the threshold by the image histogram. The scientific novelty of the proposed method is the calculation of the threshold for binary segmentation based on a histogram. The binary segmentation threshold is calculated iteratively. The studies were carried out on images of nanomaterials obtained using transmission electron microscopy. The accuracy was used to numerically evaluate the results of the proposed method. The method does not require prior training. Also, the method does not require large computing resources. The practical application of the method consists in further calculation of the particle size on images of nanomaterials for the analysis of structures.