Abstract: In view of fuzzy C-means algorithm is not considering the neighborhood information of the image, lead to the segmentation result is bad, combined with hidden Markov random field and Gaussian kernel function, put forward the nuclear space hidden Markov random field fuzzy C-means clustering algorithm. Introducte hidden Markov random field, introduce the spatial neighborhood information of pixels in the target function, the segmentation algorithm of noise robustness enhancement. The kernel function is introduced to map the nonlinear transform of the sample point to the high dimension feature space, which can enhance the anti jamming ability of the image segmentation, keep the image details. The standard image add noise to validate the performance of the improved algorithm. Both the visual effect and the peak signal to noise ratio of the segmented image show that the improved algorithm has better anti noise ability.

Key words: fuzzy C-means algorithm, hidden Markov random field, kernel function, prior probability function

摘要： 针对模糊C均值算法未考虑图像邻域信息，导致其分割效果不好的不足，结合隐马尔可夫随机场和高斯核函数，提出核空间隐马尔可夫随机场模糊C均值聚类算法。引入隐马尔可夫随机场，在目标函数中引入像素的空间邻域信息，使得分割算法对噪声鲁棒性增强；引入核函数，将样本点非线性变换映射到高维特征空间，增强图像分割的抗干扰能力，保持图像的细节信息。对标准灰度图像添加噪声，用以验证算法的性能。视觉效果及分割图像的峰值信噪比均显示，改进算法具有更好的抗噪能力。

关键词: 模糊C均值算法, 隐马尔可夫随机场, 核函数, 先验概率函数