Q-shift二元树复小波在图像融合中的应用

APPLYING Q-shift DUAL-TREE COMPLEX WAVELET TO IMAGE FUSION

复小波滤波器的构造较为复杂,采用Q-shift方法构造的二元树能有效地逼近复小波的实部和虚部特征,且具有近似的平移不变性;将二元树复小波变换用于不同传感器图像的融合,对来自不同传感器图像进行Dual-Tree CWT分解,得到2个低频子图和6个高频子图,将低频部分进行加权平均,高频部分采用最大值选取法进行融合。对融合结果的性能采用熵、均方根误差、平均梯度和相关系数进行评估,并与其它融合算法进行比较,结果表明:本融合方法优于同等环境下的其它方法。

The conformation of the complex wavelet filter is a little bit complicated,the dual-tree constituted by the method of Q-shift can effectively approach the real imaginary parts＇ characteristics of the complex wavelet and has approximate translation invariance as well.We apply the dual-tree CWT to fusing images derived from different sensors,and conduct dual-tree CWT decomposition on the images from different sensors,and get two low-frequency sub-images and six high-frequency sub-images,then we fuse low-frequency parts with the weighted average method and fuse high-frequency parts with the biggest value selection method.As to the performances of the fused result,we adopt the entropy,root mean square error,average gradient and correlation coefficient to evaluate them,and compare them with other fusion algorithms.The results show that the fusion method presented in the paper has predominance over other fusion methods in same environment.