基于小波系数区域相似度的医学图像融合*

根据小波变换原理,给出了一种基于小波系数区域相似度的医学图像融合方法.首先,以图像小波变换系数的区域均值、方差、协方差统计参量构造匹配度和加权因子,对高频子图进行区域融合; 然后,在低频部分采用绝对值取大的规则,进行图像融合;最后经过小波逆变换得到融合图像.实验证明,该方法得到的融合图像具有良好的视觉效果和量化指标,体现出更强的融合性能.

Medical image fusion based on regional similarity of wavelet coefficients

Based on expatiate wavelet transform principle, a fusion method based on regional similarity of wavelet coefficients was proposed. The regional similarity was defined by means of mean, variance and covariance of decomposition coefficients, the matching degree and adding coefficient could be gained, and then the fusion calculation of high frequency sub-image could be completed. For low frequency sub-image, abstract coefficient was adopted. Finally the fusion image was obtained by taking inverse wavelet transform. The experimental results show this fusion scheme can fuse details of input images successfully, and it can perfectly display information of the each input image, therefore using this image fusion method can get more satisfactory result than using others.