摘要
基于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。针对这些问题,提出了基于边界的多尺度域分层马尔科夫模型的图像分割算法,把边界作为可观测序列,使影像的特征场建立在一系列小波域提取的边界上,并建立基于边界的标号场MRF模型,将尺度间的交互集成在影像模型中,借助贝叶斯框架实现分割。通过测试图像和医学图像对算法进行检验,并与WMSRF算法进行比较,结果表明,算法在有效区分不同区域的同时很好的保留了边界信息,且提高了运行效率。
The segmentation algorithms based on MRF often exist boundary block effect, and have low operation efficiency by modeling the whole image. To solve the problems the image segmentation algorithm using multiscale domain hierarchical Markov model based on boundary was studied. It viewed an boundary as an observable series, the image characteristic field was built on a series of boundary extracted by wavelet transform, and the label field MRF model based on the boundary was established to integrate the scale interaction in the model, then the image segmentation was obtained by Bayesian framework. The test images and medical images were 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 the efficiency.
出处
《系统仿真学报》
CAS
CSCD
北大核心
2014年第8期1747-1751,1757,共6页
Journal of System Simulation
基金
国家自然科学基金项目(61170161
61202111
61100115)
山东省自然科学基金项目(ZR2011G0001
ZR2012FQ029
ZR2012FM008)
山东省高校科技计划项目(J12LN05)
山东省科技发展计划(2013GNC11012)
鲁东大学校基金项目(LY2010014)