Fuzzy cluster analysis for image data based on representative probability density function extracted from Resnet50 model

N Kim Ngan , V Thanh Vinh , T Hieu Thanh , N Duy Tan , H Phuc Thien , Chau Ngoc Tho , Bao Nha Linh and Vo Van Tai *

* Corresponding author (vvtai@ctu.edu.vn)

Article Sidebar

Full Text: PDF

Received: 05 Sep 2025
Revised: 27 Sep 2025
Accepted: 04 May 2026
Published: 17 Jun 2026
Title: Fuzzy cluster analysis for image data based on representative probability density function extracted from Resnet50 model
DOI: 10.22144/ctujos.2026.114
Views
34
Downloads
11
Crossref
Scopus
Google Scholar
Europe PMC
How to Cite
Ngân, N. K., Vinh, V. T., Thành, T. H., Tân, N. D., Thiện, H. P., Thơ, C. N., Linh, B. N., & Tài, V. V. (2026). Fuzzy cluster analysis for image data based on representative probability density function extracted from Resnet50 model. CTU Journal of Science, 62(3), 80-88. https://doi.org/10.22144/ctujos.2026.114
Issue
Vol. 62 No. 3 (2026)
Section
Natural Sciences

Abstract

This study develops a fuzzy clustering algorithm for image data with the following main stages. First, image features are extracted using the deep learning method ResNet50, specifically the first two convolutional blocks combined with global feature pooling. The extracted features are then normalized and transformed into representative probability density functions. Next, the appropriate number of clusters is determined through a self-updating suitable cluster function. Finally, the method determines the probability of each image belonging to a cluster. The proposed algorithm has been presented in detail with theoretical implementation steps and illustrated on a specific image dataset. Applications on several specific image datasets also demonstrate that the proposed algorithm yields reasonable and more advantageous results compared to some other popular algorithms through evaluation parameters.

Keywords: Cluster analysis, Extracted features, Image data, Probability density function

Tóm tắt

Thuật toán phân tích chùm mờ cho dữ liệu ảnh được xây dựng với những giai đoạn chính sau: đầu tiên là việc trích xuất đặc trưng của ảnh bằng phương pháp học sâu Resnet50 với hai khối tích chập đầu tiên và quá trình gộp đặc trưng toàn cục. Các đặc trưng này sau đó được chuẩn hóa và ước lượng thành hàm mật độ xác suất đại diện. Tiếp theo là quá trình xác định số chùm thích hợp dựa vào hàm tự cập nhật chùm phù hợp. Cuối cùng là phương pháp xác định xác suất thuộc vào chùm của mỗi ảnh. Thuật toán đề nghị đã được trình bày chi tiết các bước thực hiện về lý thuyết và được minh họa trên tập ảnh cụ thể. Ứng dụng trên một tập ảnh cụ thể cũng cho thấy thuật toán đề nghị cho kết quả hợp lý và thuận lợi hơn một số thuật toán phổ biến khác qua các tham số đánh giá.

Từ khóa: Dữ liệu ảnh, đặc trưng trích xuất, hàm mật độ xác suất, phân tích chùm

Article Details

Creative Commons License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

References

Bezdek, C., Ehrlich, R., & Full, W. (1984). FCM: The fuzzy c-means clustering algorithm. Computers & Geosciences, 10, 191–203. https://doi.org/10.1016/0098-3004(84)90020-7.

Chang, S. C., & Jeng, J. T. (2025). Interval generalized improved fuzzy partitions fuzzy c-means under Hausdorff distance clustering algorithm, 27, 834 – 852. https://doi.org/10.1007/s40815-024-01809-w.

Chen, T. L., & Shiu, S. Y. (2007). A new clustering algorithm based on self-updating process. Proceeding about Statistical Computing, Salt Lake City, Utah, 2034, 1–5. https://api.semanticscholar.org/CorpusID:18819116

Chen, J. H., Chang, Y. C., & Hung, W. L. (2020). A self-organizing clustering algorithm for functional data. Communications in Statistics - Simulation and Computation, 49(5), 1237–1263. https://doi.org/10.1080/03610918.2018.1494280

Chen, J. H., & Hung, W. L. (2015). An automatic clustering algorithm for probability density functions. Journal of Statistical Computation and Simulation, 85(15), 3047–3063. https://doi.org/10.1111/stan.12315.

Conze, P.-H., Andrade-Miranda, G., Singh, V. K., Jaouen, V., & Visvikis, D. (2025). Current and emerging trends in medical image segmentation with deep learning. IEEE Transactions on Radiation and Plasma Medical Sciences, 7(6), 545 - 569.
https://doi: 10.1109/TRPMS.2023.3265863

Bussa, S.K., & Boppana, N. K. (2025). Enhanced ResNet50 deep learning algorithm for classification of crack images in RCC structures. Asian Journal of Civil Engineering, 26, 3773–3784.
https://doi.org/10.1007/s42107-025-01396-7.

Hung, W. L., Chang-Chien, S. J., & Yang, M. S. (2015). An intuitive clustering algorithm for spherical data with application to extrasolar planets. Journal of Applied Statistics, 42(10), 2220–2232. https://doi.org/10.1080/02664763.2015.1023271

Hung, W. L., & Yang, J. H. (2015). Automatic clustering algorithm for fuzzy data. Journal of Applied Statistics, 42(7), 1503–1518. https://doi.org/10.1080/02664763.2014.1001326

Hung, W. L., Yang, J. H., & Shen, K. F. (2016). Self-updating clustering algorithm for interval-valued data. In 2016 IEEE International Conference on Fuzzy Systems (FUZZ-IEEE) (pp. 1494–1500).
https://doi.org/10.1109/FUZZIEEE.2016.7737867

Hung, W. L., Yang, J. H., Song, I. W., & Chang, Y. C. (2021). A modified self-updating clustering algorithm for application to dengue gene expression data. Communications in Statistics - Simulation and Computation, 50(2), 483–500. https://doi.org/10.1080/03610912018.1563149

Le, N. T. K., Le, T. H., & Vo, T. V (2023). Automatic clustering algorithm for interval data based on overlap distance. Communications in Statistics - Simulation and Computation, 52(5), 2194–2209. https://doi.org/10.1080/03610918.2021.1900248

Nguyen, T. T., Nguyen, T. T., Nguyen, K. N. T., & Vo, T. V. (2023). Balance-driven automatic clustering for probability density functions using metaheuristic optimization. International Journal of Machine Learning and Cybernetics, 14, 1063–1078.
https://doi.org/10.1007/s13042-022-01683-8

Nguyen, T. D., & Vo, T. V. (2024). Classifying for interval and applying for image based on the extracted texture feature. Granular Computing, 9(29), 1-20.
https://doi.org/10.1007/s41066-024-00450-0

Pham, D. T., & Vo, T. V. (2024). Improving fuzzy clustering model for probability density functions using the two-objective genetic algorithm. Multimedia Tools and Applications, 83, 45291-45314.
https://doi.org/10.1007/s11042-023-17217-5

Vo, T. V., & Nguyen, T. T. (2018). Similar coefficient of cluster for discrete elements. Sankhya B, The Indian Journal of Statistics, 80(1), 19 - 36.
https://doi.org/10.1007/s13571-018-0159-0

Vo, T. V., Nguyen, Y. H., & Dang, S. (2023). An automatic fuzzy clustering algorithm for discrete elements. Journal of the Operations Research Society of China, 11, 309-325. https://doi.org/10.1007/s40305-021-00388-z

Vo, T. V. (2025). Classifying for images based on the extracted color scales and representative probability density functions. Expert Systems with Applications, 291, 128456. https://doi.org/10.1016/j.eswa.2025.128456.

Zhou, J., Liu, T., & Zhu, J. (2019). Weighted adjacent matrix for k-means clustering, Multimedia Tools and Applications, 78, 33415–33434.
https://doi.org/10.1007/s11042-019-08009-x.