结合局部邻域特性和C-BEMD的图像融合方法

摘要

To conquer the weakness of existing in traditional image fusion method based on bidimensional em-pirical mode decomposition (BEMD), a novel fusion algorithm of multi-sensor images based on coordinated bi-dimensional empirical mode decomposition (C-BEMD) is proposed in this paper. Firstly, the source images are decomposed by C-BEMD to obtain the intrinsic mode function (IMF) components and residue components. Then, for the IMF components, a selection and weighted average fusion rule based on the local area energy is adopted. For the residue components, a selection and weighted average strategy based on local neighborhood visibility is presented. Finally, the fused image is obtained by performing the inverse C-BEMD on the combined coefficients. Experimental results show that the proposed approach provides superior performance over the image fusion methods based on wavelet transform, line and column crossed-used BEMD and complex empirical mode de-com-position in terms of both visual quality and objective evaluation criteria.%针对基于二维经验模态分解(BEMD)图像融合方法的不足,提出一种结合局部邻域特性和可协调二维经验模态分解(C-BEMD)的图像融合方法.为了克服BEMD应用于图像融合时存在的内蕴模函数(IMF)个数和频率不匹配问题,通过固定迭代次数和协调操作提出了适合图像融合的C-BEMD算法;然后利用C-BEMD对源图像进行分解获得IMF分量和残差分量,同时对IMF分量采用基于局部邻域能量的选择与加权平均策略,而对残差分量则采用基于局部邻域可见度的融合规则;最后将融合后的IMF分量与残差分量进行叠加,得到融合后的图像.融合仿真结果表明,该方法对于多聚焦图像、遥感图像和医学图像均可获得视觉效果佳、细节信息丰富的融合图像,优于基于行列交叉的经验模态分解和复数经验模态分解的图像融合方法.