基于可变加权参数的非监督SAR海冰图像分割

Unsupervised segmentation of SAR sea ice imagery based on variable weighted parameters

针对基于图像特征的非监督SAR海冰图像分割,提出了一种基于可变加权参数的MRF模型.用于分割的传统双分量MRF模型需要训练数据来估计所需的模型参数,因此不适合于非监督分割.通过在两个分量间引入基于函数的权重参数,解决了传统的双分量MRF无法满足非监督图像分割的问题.使用MRF模型能够自动估计模型参数并产生精确的非监督分割结果.实验证明所提出的算法能够有效的提高分割精度,且优于传统方法.

For unsupervised SAR sea ice image segmentation based on image features,an MRF model based on variable weighted parameters is proposed.The traditional two-component MRF model used for segmentation requires training data to estimate the required model parameters,so it is not suitable for unsupervised segmentation.The new approach solves the problem that the traditional two-component MRF cannot satisfy the unsupervised image segmentation by introducing the weight parameters based on the function between the two components.The proposed MRF model can be used to automatically estimate the model parameters to produce accurate unsupervised segmentation results.The experimental results show that the proposed algorithm can effectively improve the segmentation accuracy and is superior to the traditional methods.