基于Curvelet变换与一致性约束的遥感图像融合

Remote sensing image fusion based on Curvelet transform coupling consistency constraint

为解决当前遥感图像融合方案存在的块效应以及光谱扭曲等不足,在图像归一化区域方差的基础上,设计一致性约束规则的融合方法。利用变换,从多光谱图像中分解出亮度分量;将其与全色图像在快速离散Curvelet的变换下,求取高、低频系数;根据高频系数的归一化区域方差,建立一致性约束规则,对其进行融合;将不同低频系数的归一化区域能量特征进行比较,采用不同的融合方法对低频系数进行融合;基于融合后的子带系数,将其通过Curvelet与HSV逆变换,完成图像融合。通过对比实验发现,与当前遥感图像融合方法相比,所提方法不仅具有更高的融合质量,最小光谱扭曲度仅为2.291,其具备更高的融合效率,所需时耗为1.74s。

To solve the block effect and spectral distortion of current remote sensing image fusion schemes,a fast discrete Curvelet transform coupling consistency constraint rule based remote sensing image fusion algorithm was proposed.The HSV transform was used to obtain the luminance components of multispectral images.The fast discrete Curvelet transform was used to process the brightness components and panchromatic images for obtaining the high coefficients and low frequency coefficients.The consistency constraint rules were established using the normalized regional variance of the high-frequency coefficients to fuse the high-frequency coefficients.The normalized region energy characteristics of different low frequency coefficients were compared,and different fusion methods were used to fuse the low frequency coefficients according to the comparison results.The fusion image was obtained by fast discrete Curvelet inverse transform and HSV inverse transform.Experimental results show that,compared with the current remote sensing image fusion methods,the proposed method has not only higher fusion quality with the minimum spectral distortion of 2.291,but also better fusion efficiency with the time cost of 1.74 s.