A grid-based boundary sharpening clustering algorithm

To address the clustering problem for arbitrary shapes, in this paper, we propose a Grid-based Boundary Sharpening clustering algorithm called as “GBSharp”. This method is grounded in morphology and relies on two fundamental morphological operations: dilation and erosion. The main innovations of the proposed algorithm lie in two aspects. Firstly, we further introduce the concepts of inward dilation and bridge erosion based on the basic morphological operations to reduce the impact of the chain effect. Secondly, a unique indexing structure is designed specifically for non-empty cells in high dimensional space. In addition, to tackle the complex conditional judgments encountered in high-dimensional scenarios, we further utilize the inversion method for bridge-erosion operation. Experiments conducted on synthetic datasets and real-world datasets further validate the effectiveness and efficiency of the proposed algorithm.

Authors

• Lin Ma
• Qijing Yan
• Mengxia Lv
• Tiefeng Ma
• Mingchang Cheng

Keywords

• Grid-based
• Dilation
• Erosion
• Inversion method