基于区域的非下采样形态小波医学图像融合算法  

Region-based algorithm for non-sampling morphological wavelet medical image fusion

作　　者：曹义亲[1] 雷章明[1] 黄晓生[2] 

机构地区：[1]华东交通大学软件学院,南昌330013 [2]华东交通大学信息工程学院,南昌330013

出　　处：《计算机应用研究》2012年第6期2379-2381,共3页Application Research of Computers

基　　金：江西省自然科学基金资助项目(2010GZS0025);江西省科技支撑计划项目(2008J212)

摘　　要：提出了基于区域的非下采样形态小波医学图像融合算法。该算法首先将待融合的图像进行非下采样形态Haar小波分解成高频子带和低频子带,对低频子带图像直接按绝对值最大的规则进行融合,对各高频子带图像则先进行区域分割,对分割的区域根据其活跃度指数进行匹配,再对相匹配的区域按能量最大规则进行融合;最后根据融合后的低频子带及高频子带进行融合图像重构。实验结果表明,该算法在保持移不变形态小波融合方法优点的基础上,增强了融合图像的细节及亮度信息,同时有效地克服了对噪声和非精确配准敏感等缺点。This paper proposed a region-based non-downsampling morphological wavelet medical image fusion algorithm,and applied the non-downsampling morphological Haar wavelet decomposition which would result in the separation of high frequency sub-band and low frequency sub-band to be fused image.For low frequency sub-band,the image fused by means of the absolute maximum directly.While for the high sub-band,the image should be partitioned into regions which were to be used to match according to the activity index of each other,and then the matched regions would be fused under the rule of energy maximum.Finally,the reconstruction of the fused image would be performed in terms of the high and low frequency sub-band.As a result of this experiment,integrating the advantage of the morphological wavelet fusion which kept the shift-invariant,the algorithm enhances the details as well as the intensity of the fused image and overcomes the shortages such as the sensibility to noise and the non-precision alignment.

关 键 词：非下采样 形态小波变换 图像融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]