Cluster approach for auto segmentation of blast in acute leukimia blood slide images

Abstract

Image segmentation is an essential step in image analysis and is a requirement for feature extraction. One of the common segmentation techniques is using clustering algorithms. Currently, clustering algorithm has been used widely and popular in many fields such as in pattern recognition, image processing, and signal processing. These methods partition or group the objects of an image based on similarities and differences. This study proposed an automated color image segmentation technique using combination of saturation component base on HSI color space and clustering technique which are Moving K-means and Fuzzy K-means. Then, 7x7 pixels median filter was applied to remove unwanted noise after the segmentation process was completed. The comparison performance of the proposed technique was investigated. The experimental results yield a promising result for the combination of saturation component with Moving K-means clustering algorithm and 7x7 pixels median filter. © 2011 Springer-Verlag.