摘要
传统的模糊聚类图像分割算法只考虑像素点的灰度值,忽略了像素点的区域特性。虽然有些针对空间信息进行改进的聚类算法,但是仍局限于局部范围内。针对上述问题,提出了一种基于区域规则的核模糊C均值聚类的图像分割方法。该方法采用形态学和分水岭方法分别提取图像视觉轮廓和图像重建;以预划分的区域作为聚类入口,利用粒子群的全局寻优能力从小区域中搜索出较为准确的初始聚类中心,结合核函数并添加惩罚项完成图像聚类过程。实验结果表明,该图像分割算法对自然图片和合成图片能有效的分割出显著区域,同时具有较好的抗噪和规避不必要的毛躁能力。
The traditional fuzzy clustering image segmentation algorithm only considers the gray value of the pixel,ignoring the regional characteristics of the pixel. Although there are some clustering algorithms that improve spatial information,they are still limited to local scope. Aiming at the above problems,an image segmentation method based on region-based kernel fuzzy C-means clustering is proposed. The method uses morphological and watershed methods to extract the visual contour and image reconstruction respectively. The pre-divided region is used as the clustering entrance. The global optimization ability of the particle swarm is used to search for a more accurate initial clustering center from the small region. The function adds a penalty to complete the image clustering process. The experimental results show that the image segmentation algorithm can effectively segment the significant regions of natural and synthetic images,and has better anti-noise and avoid unnecessary frugality.
作者
林佳庆
林嘉炜
LIN Jiaqing;LIN Jiawei(School of Computer Science,Jiangsu University of Science and Technology,Zhenjiang 212100)
出处
《计算机与数字工程》
2022年第12期2785-2790,共6页
Computer & Digital Engineering
关键词
形态学
分水岭
模糊C均值
核函数
图像分割
morphological
watershed algorithm
fuzzy C-means
kernel function
image segmentation
