SAR图像的Gamma混合分布建模方法  

Gamma Mixture Distribution Modeling Method with SAR Images

作　　者：江金龙[1,2] 孙洪[2] 陈嘉宇[2] 查代奉[1] 

机构地区：[1]九江学院电子工程学院,九江332005 [2]武汉大学电子信息学院,武汉430072

出　　处：《信号处理》2014年第5期504-510,共7页Journal of Signal Processing

基　　金：国家自然科学基金项目(61261046)

摘　　要：随机信号的混合概率模型比单一概率模型有更多的灵活性,更适合复杂的分布建模。当前主要的混合概率模型有高斯混合模型、α分布混合模型和Gamma混合模型等。但高斯混合模型更适合随机变量对称分布的分布建模,而α混合模型参数多、算法复杂。SAR图像的像素值为非负值,且多为斜峰分布,更适合Gamma混合模型建模。仿真分析及数据测试都表明,本文提出的Gamma混合分布建模方法对SAR图像的像素统计分布具有更高的运算效率。There are more flexibilities mixtures of multi-distribution models than a single-distribution model for complicated random variables. Gaussian mixture models, alpha-stable mixture models and Gamma mixture models are main hybrid probability distribution ones. Gaussian mixture distributions are fitter for some symmetric distribu- tion random variables and alpha-stable mixture distributions for not only symmetric but also skewed, but more pa- rameters and intricate estimating algorithms are its shortcoming. Pixel values of SAR images are non-negative and skewed distributions, so fit for using Gamma mixture distribution models. Simulation analysis and data tests show that Gamma mixture distribution models have a higher operation efficiency for pixel statistical distribution of SAR image pixel distributions.

关 键 词：合成孔径雷达图像 Gamma混合分布 MARKOV CHAIN MONTE CARLO方法 高斯混合模型 α分布混合模型 

分 类 号：TN911.7[电子电信—通信与信息系统]