基于区域的图像融合方法

Region-based image fusion algorithm

针对传统像素级图像融合方法割裂像素间联系的问题,提出了一种基于区域的图像融合方法。首先对源图像进行离散小波框架分解,对其中一幅的低频子图利用模糊C均值聚类方法进行区域分割,并将区域表示线性映射到其他源图像,然后利用区域小波系数的结构相似度和区域能量构造融合规则,依据融合后的系数重构可得到融合图像。实验结果表明该方法能够有效地融合源图像信息并保持源图像特征。

A new region-based multi-resolution image fusion algorithm was presented.Firstly,the registered source images were decomposed into multiscale representations by using Discrete Wavelet Frame(DWF).Then,the fuzzy C means clustering algorithm was used to segment the low-low sub-image of one of source images and the multiscale regions representations were obtained.The corresponding regions in other images were located by linear mapping methods.After that,the fusion rules on every region were determined based on the region energy and the structural similarity of detail coefficients.Finally,the inverse DWF transform was executed to obtain the fused image.Experiment shows that the proposed algorithm is effective to integrate the information and retain the features of source images.