基于多尺度形态学滤波器的图像融合新方法

提出了一种保持图像细节和高抗噪性的图像融合新方法。这种方法首先对源图像进行多尺度形态学开闭滤波,得到源图像的低频平滑图像;然后应用多尺度Top-hat变换和Bottom-hat变换来提取小于相应尺度的图像细节特征。因为在较小的尺度特征中包含噪声颗粒的可能性较大,据此修正了Top-hat变换和Bottom-hat变换的相应系数;最后对以上两步骤得到的低频平滑图像和多尺度高频细节图像分别进行图像融合,应用形态学重建过程生成融合图像。实验表明,这种融合方法具有图像细节保持完整和噪声消除效果好的优点,处理效果优于传统的图像融合方法。

Image fusion based on multi-scale morphological filters

A novel image fusion scheme is proposed,which can preserve image details and remove image noises.Firstly,this scheme implements a multi-scale morphological open-close filter,to get the low-frequency smooth image for the source image.Secondly,the multi-scale Top-hat and Bottom-hat transforms are applied to extract the image detail features smaller than the current scale.Since the smaller scale features are more likely to contain noises,the coefficients of the Top-hat and Bottom-hat transforms are modified accordingly.Finally,an image fusion is applied for the low-frequency smooth image and the multi-scale high-frequency detail images.The fused image is formed with a morphological reconstruction process.The experiments indicate that this scheme has the merit of preserving image details and removing image noises,which is superior to the conventional image fusions.