一种改进的K-means聚类算法在图像分割中的应用

Application of Modified K-means Clustering Algorithm in Image Segmentation

K-means聚类算法是图像分割中比较常见的一种方式。它是一种无监督学习方法,能够从研究对象的特征中发现关联规则,因而具有强有力的处理方法。但是,由于该算法对噪声的敏感性K值及初始类心的不确定性,使其在图像分割中存在缺陷,于是提出了一种改进的K-means聚类算法来提高分割的效果。首先对图像进行平滑滤波处理,再根据相应条件找到特征向量作为初始类心,最后进行聚类操作。实验表明,本算法能够有效提取目标对象,提高图像分割的效果。

K-means clustering algorithm is a common way in image segmentation,and as a kind of unsupervised learning method,could discover the association rules from the characteristics of the research object,and thus is of strong processing ability.However,due to its sensitivity to the noise and uncertainty of the initial centroid,this algorithm still has some defects in the image segmentation.Based on this,a modified K-means clustering algorithm is proposed to improve the segmentation results.Firstly,the image is smoothed and filtered,and then the feature vector is found as the initial centroid according to the corresponding conditions;finally,the clustering operation is performed.Experiments show that this algorithm could effectively extract the target object and improve the effect of image segmentation.