说明：  提出了一种基于非下采样 Contourlet变换 ( NSCT)和马尔科夫随机场 (MRF)相结合的纹理图像分割算法。算法包括两 个步骤, 首先通过 NSCT实现对图像纹理特征的提取, 并使用模糊 C-均值完成对图像的初始分割 然后将初始分割结果用 MRF模型 表示, 通过贝叶斯置信传播得到图像的最终分割结果。实验结果表明, 对于纹理图像,该方法在分割错误率、 区域一致性以及边缘的 准确性方面都比传统小波变换的方法有了明显的改善。
(A tex ture i m age seg m entation a l go rithm based on comb i nati on of non-down sa m pling Contourlet transfor m ( NSCT) andM arkov rando m fi e l d model is proposed . The algorithm consists o f t wo steps . F ir st , the tex t u re f ea t ure of i m ag e is ex trac ted by NS CT, and the i m age is seg m ented i n itia lly by fuzzy c- m eans Second , the pr i m aril y seg m ented results are expressed byMRF mode, l and the fi nal seg m entati on re- s u lts are ga i ned v i a Bayes beli e f propag ati on . The exper i menta l resu lts show tha t this a l gor it hm is effecti ve fo r tex t ure i m age , it prov i desm uch better res u lts i n erro r ra te o f segm enta tion , reg ion ho m ogene ity and edge accuracy than tho se of trad iti onal w ave let transf o r m ing m ethods)