基于多尺度局部对比度和限邻域经验模式分解的遥感图像融合方法

传统二维EMD（empirical mode decomposition）算法在估算图像的均值包络面时易出现＂过冲＂现象,这使得分解的内蕴模式函数中会出现＂灰斑＂.针对这个问题,改进一种NLEMD（neighborhood limited empirical mode decomposition）算法.它基于局部均值平稳特性的方法估算图像的最佳局部均值,进而得到图像的均值包络面.该算法在克服了传统二维EMD算法的＂过冲＂现象的同时降低了时间复杂度.结合人类视觉系统对局部对比度敏感这一特性,提出一种基于多尺度局部对比度和NLEMD的遥感图像融合算法.该算法以NLEMD分解的内蕴模式函数的多尺度局部对比度为指导,对图像的内蕴模式函数进行融合处理.仿真实验表明：该融合算法能更清晰地反映融合图像中的复杂细节信息,提高了图像融合的质量.

Remote Sensing Image Fusion Based on Multi-scale Local Contrast and Neighborhood Limited Empirical Mode Decomposition

Generally,when one decomposes image using traditional 2D EMD(empirical mode decomposition) to obtain IMFs(intrinsic mode functions),there are gray spots in these IMFs,which resulting from existing overshoot in the average envelope of image.To overcome it,a NLEMD(neighborhood limited empirical mode decomposition) is improved in this paper,a stationary characteristic of local average distribution is used to acquire optimal local average of image,moreover,extracting average envelope of image.Comparing with traditional 2D EMD,it not only overcomes the overshoot,but also decreases time complexity of 2D EMD.Combination with the fact that human vision system is sensitive to local contrast,a remote sensing image fusion algorithm using multi-scale local contrast and NLEMD is presented in this paper.Fusing IMFs based on multi-scale local contrast of IMFs.Experiments show the fusion algorithm can obtain fusion image with more image sharpness and more image quality.