Spatially variant mixture multiscale autoregressive Modeling of SAR imagery for unsupervised segmentation

CASIA OpenIR > 09年以前成果

	Spatially variant mixture multiscale autoregressive Modeling of SAR imagery for unsupervised segmentation
	Ju, YW ; Tian, Z
发表期刊	CHINESE JOURNAL OF ELECTRONICS
	2006-04-01
卷号	15 期号:2 页码:359-362
文章类型	Article
摘要	A new and efficient Spatially variant mixture multiscale autoregressive (SVMMAR) model is presented for unsupervised segmentation of Synthetic aperture radar (SAR) imagery. Unlike previously proposed Bayesian segmentation approaches based on mixture model, which have used a single-resolution representation of the observed image, the proposed model can be used for the pyramid representation of SAR imagery. So, it is capable of not only describing spatially variant characteristics and complexity, but also characterizing and exploiting multiscale stochastic structure inherent in SAR imagery due to radar speckle. The estimation of parameters of the model is easily performed via Least square (LS) estimation and Expectation maximization (EM) algorithm. Then, the proposed algorithm yields maximum likelihood estimates of the labels themselves and results in unambiguous pixel labels. Moreover, a kind of method for selecting number of classes at a coarser scale is proposed, which reduced computation amount greatly. The advantage of our proposed segmentation approach is that it is not only fast, able to automatically estimate all the model parameters, and easy to implement, but also obviates the need for an explicit labeling rule and speckle reduction preprocessing before segmentation. Therefore, the model can be exploited for SAR Automatic target recognition (ATR). All of that are demonstrated by the experimental results.
关键词	Spatially Variant Mixture Multiscale Autoregressive Model Least Square Estimation Em (Expectation Maximization) Algorithm Unsupervised Segmentation Sar (Synthetic Aperture Radar) Imagery
WOS标题词	Science & Technology ; Technology
收录类别	SCI
语种	英语
WOS研究方向	Engineering
WOS类目	Engineering, Electrical & Electronic
WOS记录号	WOS:000236935300039
引用统计	被引频次：2[WOS] [WOS记录] [WOS相关记录]
文献类型	期刊论文
条目标识符	http://ir.ia.ac.cn/handle/173211/9337
专题	09年以前成果
作者单位	1.Northwestern Polytech Univ, Dept Appl Math, Xian 710072, Peoples R China 2.Chinese Acad Sci, Natl Key Lab Pattern Recognit, Beijing 100080, Peoples R China 3.Chinese Acad Sci, Inst Automat, Beijing 100080, Peoples R China
推荐引用方式 GB/T 7714	Ju, YW,Tian, Z. Spatially variant mixture multiscale autoregressive Modeling of SAR imagery for unsupervised segmentation[J]. CHINESE JOURNAL OF ELECTRONICS,2006,15(2):359-362.
APA	Ju, YW,&Tian, Z.(2006).Spatially variant mixture multiscale autoregressive Modeling of SAR imagery for unsupervised segmentation.CHINESE JOURNAL OF ELECTRONICS,15(2),359-362.
MLA	Ju, YW,et al."Spatially variant mixture multiscale autoregressive Modeling of SAR imagery for unsupervised segmentation".CHINESE JOURNAL OF ELECTRONICS 15.2(2006):359-362.