基于小波包变换的区域图像融合方法

提出了基于小波包变换的区域图像融合方法。首先对参加融合的两幅图像进行小波包多尺度分解,得到两个金字塔结构。然后,先分别提取每个金字塔最高层(分辨率最低层)的近似分量图像的边缘,把两边缘图像进行或运算以获取两幅图像的共同目标的边缘图像,并求两近似分量图像的均值产生一幅新的近似分量图像,使用文章提出的“区域聚类”的方法对新的近似分量图像进行分割产生一幅标签图像,此图中的每一个区域对应图像中的一个目标,同一区域中的所有像素都取同一个值,不同的区域取不同的值。利用标签图像,对小波包分解后最高层所有的高频细节分量图像按区域能量取大的规则进行区域对区域的融合,形成融合二值决策图。利用此二值决策图对最高层的低频近似分量图像和细节分量图像进行融合,最后重构。并采用均方根误差对该方法进行了客观评价。实验结果表明该方法有较好的融合效果,其融合性能比基于窗口能量取大的单个像素的小波包图像融合方法的融合性能好。该方法体现了真正意义上的“目标融合”。

Image Fusion Method Based on Region and Wavelet Packet

In this paper,image fusion method based on region of image and multi-resolution analysis of wavelet packet is presented.Firstly,images involved in fusion are decomposed into pyramids.A new method of image segment is proposed,the two approximations of the highest layer of the pyramids is segmented into regions to produce a Labeling image.In the Labeling image,each region is corresponding to an object and the same vale is given for the pixel of the region.Different region has different value.Secondly,a binary decision map is produced by means of Labeling image and the rule of selecting the pixel that the region energy is big.Finally,the fused image is reconstructed.The method's performance is evaluated by using Error of mean square root.The experiment results show that it has good effective.The performance is better than the performance of the fusion method based on window.It's a real 'object fusion'.