摘要
基于聚类的图像分割算法是其中常见的一种,传统聚类算法需人为确定初始聚类中心和类别数,针对如何确定最优聚类类别数的问题,基于邻接矩阵提出一种自适应图像分割算法,该算法克服了传统聚类算法人为确定初始聚类中心和聚类类别数而导致局部最优的缺陷。利用实验数据将算法和传统聚类算法比较,并应用于图像分割。实验结果显示,算法稳定性较好,能自适应的得到准确地聚类类别数,且鲁棒性较强,在应用于图像分割时的聚类结果相对与传统聚类算法更加准确。
The image segmentation algorithm based on clustering is a common one. Traditional clustering algorithm requires the determination of the initial cluster centers and cluster number of categories, and how to determine the optimal cluster number of categories is a major challenge. An adaptive image segmentation algorithm based on the adjacency matrix is proposed to overcome the local optimization caused by artificial determination of the initial cluster centers and cluster number of categories by traditional clustering algorithms. The proposed algorithm is compared with the traditional algorithm by experiment and applied to segmentation. Experimental results demonstrate good robustness and stability of the algorithm with more accurate result of clustering for segmentation than those by the tradi- tional algorithm.
出处
《电子科技》
2016年第2期66-69,共4页
Electronic Science and Technology
基金
国家自然科学基金资助项目(61202376)
上海市教育基金会晨光计划基金资助项目(10CG49)
关键词
图像处理
图像分割
聚类算法
邻接矩阵
自适应
迭代
image processing
image segmentation
clustering algorithm
adjacency matrix
adaptive
iteration
