基于Ant-Tree聚类算法的图像分割

Image segmentation based on improved Ant-Tree algorithm

图像分割可以看作对具有不同特征的像素进行聚类的过程。综合考虑像素的灰度、梯度及邻域等特征,将Ant-Tree聚类算法引入图像分割中。针对Ant-Tree算法的聚类结果信息冗余的缺点,采用了一种改进的树结构模型来提高聚类速度。此外,还提出了一种新的初始化方法,结合K-means算法动态修正聚类中心,提高了聚类准确度和算法的鲁棒性。实验结果证明改进的Ant-Tree算法可以快速准确地分割出目标,是一种非常有效的图像分割方法。

Image segmentation can be seen as the process of clustering the pixels with different characteristics, Considering the gray value, gratitude and neighborhood of the pixels synthetically, the Ant-Tree algorithm was introduced into image segmentation. As the resulting tree of Ant-Tree algorithm contains redundant information, an improved tree model was proposed in this paper. Besides, in order to optimize the process of clustering, a new initialization method was presented, and the method of K-means was also employed to modify the clustering center dynamically. Experiments and comparisons show that the Ant-Tree based clustering algorithm is an effective and efficient approach in image segmentation.