摘要: 在图像分割中,传统Markov随机场(MRF)模型基于固定形状邻域,不能完全描述边缘结构等细节信息。为此,提出一种新的MRF线型可变邻域结构,采用部分加权的先验能量模型,对图像上下文信息进行建模,将邻域的选择和图像标记相结合,通过迭代优化得到图像分割。实验结果证明,与传统模型相比,该模型能更完整地保持目标的边缘形状,在细节描述方面具有较好的结果。
关键词:
Markov随机场,
线型结构,
部分加权,
可变邻域,
邻域选择,
图像分割
Abstract: Traditional Markov Random Fields(MRF) methods used in image segmentation are based on fixed neighborhood, and cannot fully describe detailed information such as edge structure. A new linear variable neighborhood structure of MRF model is presented, and a partial weighted priori energy model is designed to model context information. Neighborhood selection and image label are combined in algorithm, using iteration to get image segmentation results. Experimental results prove that the algorithm can better keep the target edge shape, with good results in detail.
Key words:
Markov Random Fields(MRF),
linear structure,
partial weighted,
non-fixed neighborhood,
neighborhood selection,
image segmentation
中图分类号:
祁凯,吴秀清. 一种可变邻域Markov随机场图像分割模型[J]. 计算机工程.
QI Kai, WU Xiu-qing. An Image Segmentation Model of Non-fixed Neighborhood Markov Random Fields[J]. Computer Engineering.