Study of color blood image segmentation based on two-stage-improved FCM algorithm.
- Author:
Bin WANG
1
;
Huaiqing CHEN
;
Hua HUANG
;
Jie RAO
Author Information
1. Institute of Biomedical Engineering, West China Medical Center, Sichuan University, Chengdu 610041, China.
- Publication Type:Journal Article
- MeSH:
Algorithms;
Color;
Cytological Techniques;
methods;
Erythrocytes;
ultrastructure;
Humans;
Image Interpretation, Computer-Assisted;
Leukocytes;
ultrastructure
- From:
Journal of Biomedical Engineering
2006;23(2):282-286
- CountryChina
- Language:Chinese
-
Abstract:
This paper introduces a new method for color blood cell image segmentation based on FCM algorithm. By transforming the original blood microscopic image to indexed image, and by doing the colormap, a fuzzy apparoach to obviating the direct clustering of image pixel values, the quantity of data processing and analysis is enormously compressed. In accordance to the inherent features of color blood cell image, the segmentation process is divided into two stages. (1)confirming the number of clusters and initial cluster centers; (2) altering the distance measuring method by the distance weighting matrix in order to improve the clustering veracity. In this way, the problem of difficult convergence of FCM algorithm is solved, the iteration time of iterative convergence is reduced, the execution time of algarithm is decreased, and the correct segmentation of the components of color blood cell image is implemented.