本文選題：圖像分割 + 模糊聚類　； 參考：《西安電子科技大學(xué)學(xué)報》2017年06期

【摘要】：針對中智模糊C均值聚類算法抗噪能力弱的問題,提出嵌入隱馬爾科夫隨機(jī)場的中智模糊聚類分割算法.利用隱馬爾科夫隨機(jī)場描述圖像任意像素分類的先驗信息,將其與樣本分類隸屬度之間的信息散度作為正則項,嵌入現(xiàn)有中智模糊聚類目標(biāo)函數(shù);同時,將歐氏空間樣本通過核函數(shù)映射至高維空間,采用最優(yōu)化方法獲得隱馬爾科夫隨機(jī)場的核空間中智模糊聚類分割的迭代表達(dá)式.對標(biāo)準(zhǔn)的、現(xiàn)場采集的以及人工合成的3類灰度圖像添加一定強(qiáng)度的高斯和椒鹽噪聲進(jìn)行分割測試,實驗結(jié)果表明,這種分割算法相比基于隱馬爾科夫隨機(jī)場的模糊C均值聚類等分割算法的抗噪性能,有了顯著提高.
[Abstract]:In order to solve the problem of weak anti-noise ability of the middle intelligence fuzzy C-means clustering algorithm, a new algorithm of middle intelligence fuzzy clustering segmentation based on embedding hidden Markov random field is proposed. Using hidden Markov random field to describe the priori information of any pixel classification of image, the information divergence between it and the membership degree of sample classification is taken as the regular item, and the existing objective function of fuzzy clustering is embedded. The sample of Euclidean space is mapped to high dimensional space by kernel function, and the iterative expression of intelligent fuzzy clustering segmentation in kernel space of hidden Markov random field is obtained by optimization method. The standard, on-site and synthetic grayscale images with certain intensity of Gao Si and salt noise are segmented. The experimental results show that, Compared with the fuzzy C-means clustering algorithm based on Hidden Markov Random Field, the proposed segmentation algorithm has a better performance against noise.
【作者單位】： 西安郵電大學(xué)電子工程學(xué)院;
【基金】：國家自然科學(xué)基金重點資助項目(61136002) 陜西省自然科學(xué)基金資助項目(2014JM8331,2014JQ5183,2014JM8307) 陜西省教育廳科學(xué)研究計劃資助項目(2015JK1654)
【分類號】：TP391.41
，

本文編號：1886461