A Multi-Orientation Texture-Based Clustering Algorithm for Image Datasets Using GLCM and Genetic Algorithms
This study introduces an optimal clustering framework for image datasets that incorporates multi-orientation texture information from the gray-level co-occurrence matrix (GLCM) and the genetic algorithm. The proposed framework consists of three key components: (1) the extraction of multi-orientation texture features using GLCM, (2) the estimation of the optimal number of clusters through the self-updating clustering technique, and (3) the refinement of clustering assignments using the genetic algorithm, aimed at maximizing intra-cluster homogeneity and inter-cluster separability. The effectiveness of the proposed algorithm was demonstrated through a descriptive example and two specialized applications in medical image clustering. The experimental results suggest that the proposed framework provides competitive performance compared with traditional clustering techniques in terms of the Adjusted Rand Index (ARI) and computational time. Specifically, the proposed method achieved ARI values of 96.76% for the brain MRI dataset and 95.13% for the Vietnamese lung X-ray dataset. These results are slightly higher than those obtained by DEC 1 and DEC 2, which achieved ARI values of 94.91% and 93.42%, respectively.
Authors
- Tai Vovan (ORCID: https://orcid.org/0000-0002-1343-4647)
- Dinh Phamtoan (ORCID: https://orcid.org/0000-0001-7873-9794)
- Huong Nguyen Thi Diem
Institutions
- FPT University (VN)
- Can Tho University (VN)
- Van Lang University (VN)
- Jeonbuk National University (KR)
Publication Details
- Journal
- Model Assisted Statistics and Applications
- Published
- 2026-11-13
- DOI
- https://doi.org/10.1177/15741699261509310
- Primary Topic
- Advanced Clustering Algorithms Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00