Shearlet变换耦合细节强化因子的遥感图像融合算法

A remote sensing image fusion algorithm based on Non Subsampled Shearlet Transform coupling detail enhancement factor

为克服当前较多遥感图像融合方法存在间断以及吉布斯现象,本文利用像素点间灰度以及梯度信息,设计了一种采用非下采样Shearlet变换(NSST)耦合细节强化因子的图像融合方法。将多光谱(MS)图像经过强度-色调-饱和度(IHS)变换,分离出强度成分。随后,借助NSST变换处理强度成分与全色(PAN)图像,获取对应的高频和低频系数。以强度成分对应的低频系数为依据,通过图像的空间频率特性计算加权系数,将PAN图像的低频系数植入到强度(I)成分对应的低频系数中,融合低频系数。采用像素点间灰度以及梯度信息,构造细节强化因子,融合高频系数。最后,采用IHS和NSST反变换重构这些融合系数,获取融合结果。实验结果显示:较当前融合技术,所提算法拥有更为理想的融合效果,具有更高的互信息值和更低的光谱偏差度值。

In order to overcome the discontinuities and Gibbs phenomenon in many remote sensing image fusion methods,this paper designs an image with Non Sampling Shearlet Transform(NSST)coupling detail enhancement factor by using the gray level and gradient information between pixels fusion method.The intensity(I)component of Multi Spectral(MS)image is separated by Intensity-Hue-Saturation(IHS)transformation.The high and low frequency coefficients of I component and Panchromatic(PAN)image are extracted by NSST.Based on the low-frequency coefficient corresponding to component I,the filling coefficient is calculated by the spatial frequency characteristics of the image.The low-frequency coefficient corresponding to the image is filled into the low-frequency coefficient corresponding to component I,and the low-frequency coefficient is fused.The gray level and gradient information between pixels are utilized to construct detail enhancement factors to measure the differences between pixels and their neighbors,and then the high-frequency coefficients are fused.Based on the fusion coefficient,IHS and NSST inverse transforms are adopted to reconstruct the coefficients,and the fusion results are obtained.The experimental results show that the image fusion algorithm has higher mutual information value,lower spectral deviation value and better spectral and spatial characteristics than the current image fusion algorithm.