基于贝叶斯框架和Gamma分布的SAR图像分割

SAR Image Segmentation Based on Bayesian Framework and Gamma Distribution

图像分割作为图像处理的关键步骤之一,一种高精度的SAR图像分割方法在图像分析与解译中显得尤为重要。根据SAR图像的统计特性,本文主要研究基于贝叶斯框架和Gamma分布的SAR图像分割方法,分别以单一Gamma分布概率模型和Gamma混合模型定义像素属于聚类的或然率;此外,基于马尔科夫随机场理论定义像素与其邻域像素标号间相互作用关系,并以此作为先验概率;然后,根据贝叶斯定理得到像素属于聚类的后验概率。最后,分别基于上述两种方法对模拟SAR图像及真实SAR图像进行分割实验。结果表明,Gamma混合模型较单一Gamma分布能够更加准确地描述SAR图像分布特征,得到高精度的分割结果。

Image segmentation is one of the key steps of image processing.A high-precision SAR image segmentation method is very important in image analysis and interpretation.According to the statistical properties of SAR images,this paper mainly studies the SAR image segmentation method based on Bayesian framework and Gamma distribution.The probability of the pixels belonging to clustering is defined by the single Gamma distribution model and the Gamma Mixture Model (GaMM).In addition,the Markov Random Field (MRF) is used to define the the prior probability to describe the interaction of the label between the pixel and its neighborhood pixels.Then the posterior probability is obtained by Bayesian theorem.Finally,the simulated SAR images and real SAR images are segmented experimentally based on the above two methods.The experimental results show that the Gamma mixture model can describe the image distribution more accurately than the single Gamma distribution,and obtain higher precision segmentation results.