结合边界的小波域马尔科夫模型的图像分割算法  

Image segmentation algorithm of wavelet domain Markov model with boundary

作　　者：唐文静[1] 许兆新[2] 

机构地区：[1]鲁东大学信息与电气工程学院,山东烟台264025 [2]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《传感器与微系统》2014年第9期145-147,154,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61170161);山东省自然科学基金资助项目(ZR2012FQ029;ZR2012FM008);山东省高校科技计划资助项目(J12LN05);山东省科技发展计划资助项目(2013GNC11012);鲁东大学校基金资助项目(LY2010014)

摘　　要：针对基于马尔科夫随机场(MRF)的分割算法常存在边界块效应,且对整幅图像进行建模运行效率低等问题,提出了结合边界的小波域马尔科夫模型的图像分割算法,把影像的特征场建立在一系列小波域提取的边界上,并建立相应的边界标号场MRF模型,借助贝叶斯框架和SMAP准则实现分割。利用Matlab GUI实现了分割系统,通过医学图像检验,结果表明:相比于小波域分层随机场模型(WMSRF),该算法在有效区分不同区域的同时很好地保留了边界信息,提高了运行效率。Aiming at problems that segmentation algorithms based on MRF often occur boundary block effect and low operation efficiency by modeling the whole image, image segmentation algorithm using wavelet domain Markov model with boundary is proposed. In which the image characteristic field is built on a series of boundary extracted by wavelet domain, and corresponding label field MRF model with boundary is established, then image segmentation is obtained by Bayesian framework and SMAP. The segmentation system based on Matlab GUI is implemented, and the medical images are utilized to test the proposed algorithm. The results show that it can not only distinguish effectively different regions, but also retain the boundary information very well, and improve ooeration efficiency comoared with WMSRF.

关 键 词：图像分割 小波域 马尔科夫 边界 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]