摘要： 提出一种基于邻域特征和聚类的图像分割方法。该方法提取像素点的5维邻域特征，利用渐进聚类方法使同类元素具有较高的相似度、不同类元素相似度差别较大，从而对图像中的像素进行归类划分，实现目标图像的正确分割。实验结果表明，该方法能准确定位图像边缘，具有较强的抗噪性和较高的分割精度。

关键词: 图像分割, 邻域特征, 渐进聚类, 模糊C均值

Abstract: This paper presents an image segmentation method based on neighborhood features and clustering. In the method, five dimensional neighborhood features of pixels are extracted, incremental clustering is used in order to make similar elements have a high similarity and different elements have a vary greatly different similarity, pixels in the image are classified division to achieve the correct segmentation of target image. Experimental results demonstrate that this method can accurately locate image edge, it has a strong anti-noise property and higher accuracy of segmentation.

Key words: image segmentation, neighborhood feature, incremental clustering, Fuzzy C-Mean(FCM)

中图分类号: 

• TP391