Development of an Image Clustering Algorithm for Nanoparticle Extraction

The article explores the application of the k-means method and its improved versions for clustering nanomaterial images. The purpose of the study is to develop an effective clustering method capable of identifying nanomaterial particles in images with high noise levels. The standard k-means method is compared with its modifications, including as k-means++, MiniBatch k-means, and Elkan, to optimize pixel allocation. The unique properties of these modified methods, including their approaches to assigning pixels to centroids and forming clusters, are examined in detail. Noise reduction filters are described as a preprocessing step to enhance clustering. Practical recommendations for optimizing the clustering process are provided, which can improve accuracy, reliability, and noise resilience. The proposed method is based on clustering techniques that partition images into regions corresponding to the average intensity values within those regions.