基于Contourlet的医学图像融合技术探讨

Study of a medical image fusion algorithm based on Contourlet regional statistics

针对不同医学影像设备获得的多源图像信息有效融合和综合利用的问题,提出一种基于Contourlet区域特性的医学图像融合算法———CRSIF算法,借助于Contourlet变换的优良特性,在Contourlet变换域使用加权平均和选择方式实现频域系数的有效融合,对低频子带采用局部加权能量作为评价标准,高频子带采用区域加权Contourlet对比度的树型结构设计,以满足区域的融合规则。对CT/MR脑部医学图像的仿真分析表明,该算法可克服传统规则下融合图像不连续及产生毛刺和斑点的缺陷,使融合图像与人类视觉系统的感知特性相吻合。

In order to effectively integrate and comprehensively utilize medical images from multiple sources, a medical image fusion algorithm based on contourlet reginal statistics （CRSlF） was proposed. Taking full advantage of contourlet transformation excellent characters, the weighted average and selective mode were used to integrate frequency domain coef ficient in contourlet transform domain. Local weighted energy and regional contourlet weighted contrast were taken as the evaluation standards in low sub-band and high band, respectively, in order to satisfy the fusion algorithm. Extensive fusion experiments on CT and MR medical images showed that the proposed approach could effectively ensure the fusion images match human vision perceptual characteristics, providing a more satisfactory outcome than conventional image fusion algo rithms.