摘要
由于遥感图像中存在边缘混叠和斑点噪声等问题,基于经典的马尔科夫随机场模型的分割算法效果并不太理想。本文针对遥感图像分割中某些像素分类的不确定性,将模糊MRF模型引入遥感图像分割领域,构造了基于MAP-FMRF的分割框架,提出了新的模糊MLL模型,在充分结合空间信息的同时利用灰度和纹理共同构造特征空间,以修正的EM算法结合SA算法获取全局最优解,实现无监督分割,实验对比证明该方法准确率更高。此外,文中还设计了一种新的优化方案以提高分割的效率。
Because remote-sensing applied, so the segmentation result image has interlaced edges and speckle noise and the standard MRF can not be of remote sense images using standard MRF is not satisfied. Aiming at the ambiguity in the segmentation of remote sense images, we apply the fuzzy Markov random field model to the segmentation of remote sense images. In this paper, a framework based on MAP-FMRF and a new fuzzy MLL model are proposed. We use gray and texture to construct the eigenspace with space information. Then we apply the EM algorithm and simulated annealing to search global optimal resolutions to realize the unsupervised image segmentation. Experiment comparison shows that our method is more efficient. In addition, we also devise an optimized scheme to improve the efficiency of segmentation.
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2009年第1期213-218,共6页
Chinese Journal of Scientific Instrument