基于机器学习SAR图像噪声抑制技术研究  

Research on Machine Learning Based SAR Image Noise Blanking Technology

作　　者：徐东[1] 张晓雯[1] 

机构地区：[1]海军大连舰艇学院基础部,辽宁大连116018

出　　处：《软件导刊》2013年第10期163-165,共3页Software Guide

摘　　要：为提高含噪声SAR(合成孔径雷达)图像的目标识别能力,提出基于机器学习的图像噪声抑制技术的研究思路。该思路通过问题估计,为含噪声图像获得清晰的输出景物,从而生成一个与含噪声图像相符合的增强的合成景物世界。该技术以马尔可夫网络为体系,根据图像与景物、景物与景物之间的联系来确定网络上的信息传递规则,从大量的训练事例中学习并获得网络参数,利用贝叶斯方法为原图像找到理想的后验概率,生成一个清晰的超分辨率结果图。This paper suggested an idea of machine leaning based image noise blanking technology for improving the ability target recognition of SAR noise image, the idea was to obtain a clear scene for the noise image , through the problems- es- timating , It could generate a compatible with noise image and enhanced synthesis of scene image ~ The technology mod- eled that world with a Markov network which learned from a number of training examples and then got the network pa- rameters according to image and scene, scene and scene to determine the link between the messages on the network rules. Bayesian method allowed it efficiently find a fitting posterior probability for the scene, generating a clear high resolution images.

关 键 词：SAR图像 马尔可夫网络 贝叶斯 噪声抑制 

分 类 号：TP317.4[自动化与计算机技术—计算机软件与理论]