对遥感图像经Contourlet变换后的高频子带系数分布的方向特征进行统计分析, 发现遥感图像经Contourlet变换后高频系数的分布具有较强的方向区域特征, 在此基础上, 提出一种基于Contourlet系数方向区域相关性的遥感图像融合算法, 该算法首先对多光谱图像经IHS变换后的亮度分量和全色图像分别进行Contourlet变换, 然后以多光谱图像亮度分量的低频信息作为融合图像亮度分量的低频信息, 通过计算并比较全色图像的高频系数和对应的多光谱图像亮度分量的高频系数的方向区域匹配度确定融合图像亮度分量的高频信息; 最后经过Contourlet逆变换和IHS逆变换获得融合图像。实验结果表明, 该算法在提高融合图像空间分辨率的同时能够更好地保留原始多光谱图像的光谱信息, 与传统遥感图像融合算法相比, 该算法具有较好的融合图像信息熵和清晰度, 具有一定的实用性。

In this paper, we analyze the directional characteristics of contourlet coefficients in high frequency subbands of remote sensing images. And then we find that each coefficient in these subbands has obvious directionality. Based on this directional characteristic, this paper proposes a novel image fusion algorithm for remote sensing images. First, we separately perform contourlet transform on the intensity component of the multi-spectral remote sensing image obtained by IHS transform, and the panchromatic remote sensing image. Second, we choose the low frequency coefficients of multi-spectral image’s intensity component to form the low frequency subband of target image. Subsequently, we compare the directional matching degree of the high frequency coefficients of the panchromatic image with those of the multi-spectral image’s intensity component to determine the high frequency subbands of target image. Finally, the target image is obtained by inverse contourlet transform and inverse IHS transform. Extensive experimental results show that the proposed method is superior to conventional methods in terms of entropy, joint entropy, and average gradient. It can enhance the spatial resolution of target images. Meanwhile, it well preserves the color information of multi-spectral images.