基于简化随机场模型的高分辨率遥感影像分割方法

Remote Sensing Image Segmentation Algorithm Based on Simplified Random Field Model

提出了一种灰度分割的基础上添加辅助的纹理分割的基于简化随机场模型的遥感影像目标分割方法,即用常用的描述局部图像特点的特征代替MRF中定义的特征,将这些特征组合成特征向量进行模糊C均值聚类完成分割。给出了算法流程和实验结果,并将该结果与基于高斯马尔可夫随机场模型法分割的结果进行比较,实验结果表明简化随机场模型法在保证一定的分割精度的情况下,分割速度明显快于高斯马尔可夫随机场模型法。

Taking into account both gray-value and texture feature,thi s paper proposes a remote sensing image segmen-tation algorithm based on simplifi ed Random Field Model.Firstly,some common features that describe local image substi-tute the features of MRF and form a eigenvector.Then the eigenvector is used to perform Fuzzy C-Mean(FCM)clustering and the segmented image is achie ved.In addition,this paper gives the algorithm flow and tests the method with 1m IKONOS image.The results shows that this algorithm has faster computing sp eed and better region integrality than stan-dard GMRF.