当前较多遥感图像融合算法是利用主成分分析方法来完成遥感图像的融合, 由于主成分分析方法融合后的图像会产生光谱畸变, 易导致所融合图像存在光谱失真的问题。对此, 设计了一种采用双正交小波变换耦合区域梯度特征的遥感图像融合算法。对多光谱图像进行色调-饱和度-亮度变换, 以获取多光谱图像的亮度分量, 引入双正交小波变换将该亮度分量与全色图像进行小波域分解, 以获取图像的低频与高频子带; 通过低频子带中像素点的区域梯度特征构造均值梯度模型, 用于求取低频子带融合系数, 利用高频子带中像素点对应的区域方差构造相似度因子, 用于求取高频子带融合系数; 通过色调-饱和度-亮度与双正交小波的逆变换获取所融合遥感图像。仿真实验结果显示, 所设计方法与当前遥感图像融合方法相比, 融合的遥感图像具有更好的视觉效果。

The current many remote sensing image fusion algorithms use orthogonal filter to obtain the high frequency and low frequency bands of the image and then complete the fusion of remote sensing images. Because the orthogonal filter can produce the phase distortion, it can easily lead to the defects of the fused image, such as the edge distortion. Therefore, a remote sensing image fusion algorithm based on biorthogonal wavelet transform coupled with regional gradient features is designed. First, the HSV transform of multispectral images is used to get the luminance components of multispectral images. The biorthogonal wavelet transform is introduced to decompose the luminance component and panchromatic image in the wavelet domain, so as to get the low frequency and high frequency sub-band of the image. Then, the mean gradient model is constructed through the gradient characteristics of the pixels in the low frequency sub-band, which is used to obtain the low frequency sub-band fusion coefficient. The similarity factor is constructed by using the region variance corresponding to pixels in the high frequency sub-band, which is used to obtain the high frequency sub-band fusion coefficient. Finally, the fused remote sensing image is got by inverse transformation of HSV and biorthogonal wavelet. The simulation results show that the proposed method has better visual effect compared with the current remote sensing image fusion method.