带边缘惩罚和自适应权马尔科夫随机场的合成孔径雷达图像快速分割

Efficient Segmentation of SAR Images Using Markov Random Field Models with Edge Penalties and an Adaptive Weighting Parameter

针对合成孔径雷达(SAR)图像提出了一种带边缘惩罚和自适应权马尔科夫随机场(MRF)模型的快速分割算法。在MRF分割模型的能量函数中引入了边缘惩罚和自适应加权参数。边缘惩罚的引入能够减少边缘模糊从而更加精确地定位边缘。自适应权参数能够根据图像分割时收敛的阶段以及图像的局部场景自适应地调整能量函数中数据模型因子的权重,这有利于在均质区域改进分割区域的一致性,而在非均质区域则可保持图像的边缘和重要细节。针对所提出的能量函数提出了一种快速的非均质点跟踪优化算法。对合成和真实的SAR图像的分割结果表明,所提出的分割方法能提高分割的精度并显著减少运行时间。

An efficient synthetic aperture radar （SAR） image segmentation approach using a Markov random field （MRF） model with edge penalties and an adaptive weighting parameter is proposed. The edge penalty and the adaptive weighting parameter are introduced into the energy function of MRF segmentation model. As a result, the edge fuzzy is reduced with the introduction of edge penalty. The adaptive weighting parameter can adjust adaptively the weights of the data modeling factor in the energy function according to the stages of iteration and the heterogeneity of local scenes, which is in favor to get smoother results for homogeneous regions and preserve edges and important image details for heterogeneous regions. An efficient optimization algorithm called heterogeneous point tracking algorithm is presented in terms of the characteristics of the energy function. Experiments with simulated data and real SAR images show that the proposed algorithm improves the segmentation accuracy and reduces the running time.