基于减法聚类和K均值聚类的彩色图像分割算法

Region-Based Image Segmentation with Improved Subtractive Clustering and K-means Clustering

传统图像分割方法大都存在分割速度低下、过度分割等缺点.针对上述问题,提出一种新的彩色图像区域分割算法.这种方法首先将图像转化至L*a*b*空间,并划分为子块,抽取图像子块的颜色、纹理和位置特征组成子块的特征向量,然后运用减法聚类,获得聚类簇数和初始蔟中心,最后利用改进的K均值算法在像素点特征空间进行聚类,进而分割图像成区域.实验结果表明这种新方法具有分割效率高、分割效果理想等优点.

The traditional image segmentation methods are often of some disadvantages, such as the lower segmentation speed, the over segmentation etc. To resolve these problems, a new region-based color image segmentation algorithm is introduced in the paper, which transforms an image to L＊a＊b＊ space, and it divides the image into sub-blocks, and it extracts color, texture and position features for every sub-block firstly. Then, the number of cluster and the initial cluster center are obtained by using subtractive clustering algorithm. Finally, the cluster program is completed by the application of a modified K-mean algorithm in the feature space of pixel. The experimental results show that this new method has the advantages of high speed of image segmentation, good segmentation effect etc.