小波域正则化遥感图像融合

遥感图像融合过程中,为了在增强空间分辨率的同时减少光谱损失,文章提出了一种改进的强度-色调-饱和度和小波的融合算法,新算法在原算法中,引入正则化技术。以小波域局部高斯模型作为光谱分布先验概率,以全色图小波系数做为空间分布先验概率,以马尔可夫随机场描述空间特征,通过梯度下降法迭代优化,实现了光谱和空间信息总损失最小的图像融合。对美地球资源卫星5的增强专题图像的融合试验证明文中提出的算法可同时提高多光谱及全色图像与融合图像的相关性,有效改善融合效果。

Improving Remote Sensing Image Fusion Based on Regularization in Wavelet Domain

Aim.Ref.4,authored by M.Choi,should and can,in our opinion,be further improved.In the full paper,we explain our improvements in some detail;in this abstract,we just add some pertinent remarks to naming the first two sections.Section 1 is: image fusion based on regularization in wavelet domain.In section 1,we present the regularization conditions as shown in eq.(11) deduced by us.Section 2 is: the fusion algorithm and its implementation.Its three subsections are: the algorithm(subsection 2.1),the rules for fusing wavelet coefficients(subsection 2.2) and the seven-step procedure of our new algorithm.In subsection 2.1,we use the gradient descent technique to iterate the wavelet domain,thus accomplishing the image fusion with the least loss of spectral information and spatial characteristics.In subsection 2.2,we use eq.(16) to compute the fusion coefficients of multi-spectral and panchromatic images.The analysis and comparison of experimental results and statistical results,shown in Fig.1 and Table 1 respectively and obtained by utilizing the U.S.Landsat-7 Enhanced Thematic Mapper Plus(ETM ) data,point out preliminarily that our algorithm can effectively enhance the spatial characteristics of images and preserve their spectral information.