基于广义多分辨似然比的SAR图像无监督分割

计算机工程与应用 ›› 2007, Vol. 43 ›› Issue (11): 4-6.

基于广义多分辨似然比的SAR图像无监督分割

张前进郭雷

西北工业大学西北工业大学

收稿日期:2006-10-05 修回日期:1900-01-01 出版日期:2007-04-11 发布日期:2007-04-11
通讯作者: 张前进

Multiresolution Likelihood Ratio for Unsupervised Segmentation of SAR Imagery

Received:2006-10-05 Revised:1900-01-01 Online:2007-04-11 Published:2007-04-11

摘要/Abstract

摘要： 首先在多分辨四叉树上定义了一个广义的多分辨似然比，刻画并且累积了SAR(synthetic aperture radar)图像中目标与背景在不同分辨率上的差异，从而增大了目标与背景之间的区分度。为了达到图像无监督分割目的，利用经典的混和模型方法分别估计出每个分辨率上对应的多分辨似然比中一组密度函数的参数。为了考虑被分类像素与周围像素之间的Markov性，减弱对噪声的敏感性，利用开窗内像素的广义多分辨似然值的和的大小来确定中心像素点的类别。实验中与通常的分割技术作了比较，也表明该方法不论从分割的精度，对噪声的敏感度，还是从边缘的光滑度都表明该方法都具有明显的效果。

关键词: 广义多分辨似然比, 经典混合模型, 无监督分割

Abstract: A generalized likelihood ratio test function—multiresolution likelihood ratio is defined, which cumulates the differences of background and targets at different multiresolution of synthetic aperture radar (SAR) imagery. So the multiresolution likelihood ratio increases the distinction of background and targets. In order to get unsupervised segmentation, classical mixture model is proposed and applied to estimate the parameters of multiresolution likelihood ratio. Finally we classify each individual pixel based on a test window. The method avoids some drawbacks that existed in some popular segmentation techniques. Experimental results demonstrate that our method performs fairly well.

Key words: generalized multiresolution likelihood ratio (GMLR), classical mixture model, unsupervised segmentation

张前进郭雷. 基于广义多分辨似然比的SAR图像无监督分割[J]. 计算机工程与应用, 2007, 43(11): 4-6.