光子学报, 2018, 47 (2): 0210002, 网络出版: 2018-01-30
基于最小Hausdorff距离和NSST的遥感图像融合
Remote Sensing Image Fusion Based on Minimum Hausdorff Distance and Nonsampled Shearlet Transform
遥感图像融合 最小Hausdorff距离 非下采样剪切波变换 主成分分析 稀疏表示 Remote sensing image fusion Minimum Hausdorff distance Nonsubsampled Shearlet transform Principal component analysis Sparse representation
摘要
为了最大限度地保留多光谱图像的光谱特性和全色图像的空间细节, 提出基于最小Hausdorff距离和非下采样剪切波变换(NSST)的遥感图像融合方法.首先, 将原多光谱图像进行主成分分析(PCA)获得其第一主分量, 选择NSST对第一主分量和全色图像分别进行分解, 得到相应的低频子带系数和高频子带系数.其次, 对低频子带系数采用基于稀疏表示的融合策略, 稀疏系数与区域空间频率相结合, 根据区域空间频率选择权值, 对稀疏系数进行加权; 对于高频子带系数充分考虑其邻域系数相关性, 提出采用最小Hausdorff距离表征相应区域相关性, 根据相关性不同采用不同的融合策略.最后, 对融合系数进行NSST逆变换得到融合后的第一主分量, 再将新的第一主分量与其他高阶主分量进行PCA逆变换得到融合图像.选择三组QuickBird卫星图像和一组SPOT卫星图像进行测试, 与传统的融合策略算法相比, 本文方法获得的融合结果客观评价指标更优, 且主观视觉效果更好.
Abstract
In order to preserve both spectral and spatial information simultaneously in fused image, we introduce the minimum Hausdorff distance and NonSampled Shearlet Transform (NSST) to construct a new method for remote sensing image fusion. Firstly, Principal Component Analysis (PCA) transform is applied in the original multispectral image to obtain the first principal component, this component and the panchromatic image are decomposed by NSST respectively to obtain the corresponding low frequency subband coefficients and high frequency subband coefficients. Then, the low frequency subband coefficients are fused by sparse representation, the sparse coefficients of sparse representation are fused with the region space frequency; for the high frequency subband coefficients, the regional structure similarity is utilized, using the minimum Hausdorff distance to represent the correlation of regions and different fusion strategies are adopted according to the correlation. Finally, the fused coefficients are transformed by inverse NSST to obtain the new principal component, the new component and other higher order principal components are transformed by inverse PCA transform to obtain the fused image. In this paper, three QuickBird satellite images and one SPOT satellite image are selected for testing, the results show that compared with the traditional fusion strategy algorithms, the fusion results obtained by proposed method have better objective evaluation index and subjective visual effect.
武晓焱, 柴晶, 刘帆, 陈泽华. 基于最小Hausdorff距离和NSST的遥感图像融合[J]. 光子学报, 2018, 47(2): 0210002. WU Xiaoyan, CHAI Jing, LIU Fan, CHEN Zehua. Remote Sensing Image Fusion Based on Minimum Hausdorff Distance and Nonsampled Shearlet Transform[J]. ACTA PHOTONICA SINICA, 2018, 47(2): 0210002.