發(fā)布時間：2018-03-07 09:37

  本文選題：極化SAR　切入點：相干斑抑制　出處：《遙感學報》2017年02期 　論文類型：期刊論文

【摘要】：極化合成孔徑雷達(SAR)圖像受相干斑噪聲的影響,難以很好地保持結(jié)構(gòu)特性,針對這個問題提出了一種采用3維塊匹配小波變換的非局部均值濾波算法NL-3DWT(Nonlocal Filter based on 3-D Patch Matching Wavelet Transform)。該算法使用塊匹配的3維非抽樣小波變換對極化總功率圖進行預濾波,在此基礎(chǔ)上使用邊界對齊窗提取結(jié)構(gòu)相似像素,同時使用Sigma范圍選擇極化SAR數(shù)據(jù)的散射相似像素,共同構(gòu)成相似像素集合;構(gòu)建結(jié)構(gòu)保持權(quán)重函數(shù)增大圖像結(jié)構(gòu)信息在塊相似性度量時的權(quán)重,最終實現(xiàn)極化SAR圖像結(jié)構(gòu)保持的相干斑抑制。該算法增強了圖像結(jié)構(gòu)特征的表達,提高了結(jié)構(gòu)相似像素選擇的準確性,機載極化SAR數(shù)據(jù)實驗結(jié)果表明,NL-3DWT算法能夠在抑制相干斑噪聲的同時,更有效地保持極化SAR圖像的結(jié)構(gòu)特性和極化散射特性。
[Abstract]:Polarimetric synthetic Aperture Radar (SAR) images are hard to maintain structural characteristics due to speckle noise. To solve this problem, a non-local mean filter algorithm, NL-3DWT(Nonlocal Filter based on 3-D Patch Matching Wavelet transform, is proposed, which uses block-matched 3D non-sampling wavelet transform to pre-filter the total polarimetric power map. On this basis, we use the boundary alignment window to extract the similar pixels with similar structure, and use the Sigma range to select the scattered similar pixels of the polarized SAR data, so as to form a set of similar pixels. A structure-preserving weight function is constructed to increase the weight of image structure information in block similarity measurement, and finally the speckle suppression of polarimetric SAR image structure retention is realized. This algorithm enhances the expression of image structure features. The experimental results of airborne polarimetric SAR data show that the NL-3DWT algorithm can suppress speckle noise and maintain the structure and polarization scattering characteristics of polarimetric SAR images more effectively.
【作者單位】： 合肥工業(yè)大學計算機與信息學院;
【基金】：國家自然科學基金(編號:61371154,61271381,61503111) 安徽省自然科學基金(編號:1608085QF142) 中央高校基本科研業(yè)務(wù)費專項資金資助(編號:2015HGBZ0106,2015HGQC0005) 國家博士后科學基金(編號:2016M592045)~~
【分類號】：TN957.52

【相似文獻】

相關(guān)期刊論文 前10條

1 郭景峰,苑煒_",姜愛群;基于提升方案的小波變換原理與C++實現(xiàn)[J];計算機工程與科學;2004年01期

2 王波，

本文編號：1578881