Int J Performability Eng ›› 2020, Vol. 16 ›› Issue (3): 454-461.doi: 10.23940/ijpe.20.03.p15.454461

Previous Articles     Next Articles

LDKM: An Improved K-Means Algorithm with Linear Fitting Density Peak

Chulei Zhanga,*, Honghua Cuib, Yizhang Wangc, Tiantian Zhaoc, and You Zhouc,*   

  1. aSchool of Media and Communication, College of Humanities and Science, Northeast Normal University, Changchun, 130117, China;
    bSecond Hospital, Jilin University, Changchun, 130012, China;
    cCollege of Computer Science and Technology, Jilin University, Changchun, 130012, China
  • Submitted on ; Revised on ; Accepted on
  • Contact: Chulei Zhang,You Zhou E-mail:95292043@qq.com,zyou@jlu.edu.cn

Abstract: The biggest drawback of the K-means algorithm is that the number of clusters must be specified before use, and the central point is randomly initialized. To make up for this shortcoming, this paper proposes an improved algorithm of K-means called the linear fitting density peak K-means algorithm (LDKM), which realizes the automatic initialization of K-means and improves the accuracy of the algorithm. The LDKM algorithm is applied to the field of image segmentation and compared with the K-means algorithm, and the experimental results have clear outline and less noise. The LDKM algorithm is applied to the classification and recognition of white blood cells, and the experimental results show that the LDKM algorithm can extract white blood cells completely and obtain pure results.

Key words: automatic initialization, K-means, density peak, image segmentation