基于Canopy聚类的噪声自适应模糊C-均值算法

针对局部空间信息的模糊C-均值算法（WFLICM）中空间影响因子容易受到噪声影响出现错误标识的问题，提出一种融合局部和非局部空间信息的模糊C-均值聚类图像分割算法(NLWFLICM)，在WFLICM算法的模糊影响因子中引入非局部空间信息，根据噪声程度自适应地设置局部和非局部信息权重，并重新标记中心点的模糊影响因子。实验结果表明，NLWFLICM算法具有比WFLICM算法更强的鲁棒性和自适应性，并在一定程度上提高了WFLICM算法对含有大量噪声图像进行分割的鲁棒性，同时保留了图像的纹理。为了提高算法的聚类性能和收敛速度，结合Canopy算法能够快速对数据进行粗聚类的优点，提出基于Canopy聚类与非局部空间信息的FCM图像分割改进算法（Canopy-NLWFLICM），可以在NLWFLICM算法聚类前，对聚类中心进行预处理，从而提高收敛速度和图像分割精度。

Improved fuzzy C-means clustering algorithm based on canopy clustering

Aiming at the problem that the spatial influence factors are easily misidentified by noise in the fuzzy C-means algorithm (WFLICM) for local spatial information, this paper proposed a fuzzy C-means clustering algorithm for image segmentation (NLWFLICM) based on local and non-local spatial information. It introduced the non-local spatial information into the fuzzy influencing factor of WFLICM algorithm, the weight of local and non-local information is adaptively set according to the noise level, and the fuzzy influence factors of the central point are re-marked. The experimental results show that the NLWFLICM algorithm is more robust and adaptive than the WFLICM algorithm, and improves the robustness of the WFLICM algorithm to a large extent, while preserving the image texture. In order to improve the clustering performance and convergence speed of the algorithm, combined with the advantages of Canopy algorithm for fast clustering of data, this paper proposes an improved algorithm for FCM image segmentation based on Canopy clustering and non-local spatial information (Canopy-NLWFLICM) before clustering algorithm. This can improve the convergence speed and image segmentation accuracy. Key words: clustering algorithm ; Canopy algorithm ; fuzzy C-means cluster; local and non-local spatial information