基于NSCT的多模态医学图像融合算法的研究  

A Novel Method for Multimodal Medical Image Fusion Based on Non-Subsampled Contourlet Transform

作　　者：徐磊[1] 曹艳[1] XU Lei;CAO Yan(Department of Nuclear Medicine, Nanjing First Hospital, Nanjing Medical University, Nanjing Jiangsu 210006, China)

机构地区：[1]南京医科大学附属南京医院(南京市第一医院)核医学科,江苏南京210006

出　　处：《中国医疗设备》2017年第12期63-67,共5页China Medical Devices

基　　金：国家自然科学基金(81271604)

摘　　要：目的提出一种新颖的非下采样轮廓波变换算法,并将其用于多模态医学图像融合领域。方法第一步,采用非下采样轮廓波变换将源图像分解为高频和低频子图像;第二步,低频部分采用区域均值取大融合规则,能保留图像的绝大部分信息,获得更高的对比度和清晰度,高频图像采用取大区域方差取大融合原则,能有效地突出图像细节信息;第三步,采用非下采样轮廓波逆变换将融合子图像进行重构运算,获得最终的融合图像。结果 6组医学图像融合实验表明,基于本文提出算法获得的融合图像质量最佳,且定量评价指标熵、平均值、标准差、边缘强度分别较其他算法提升0~40%、3%~42%、1%~42%和0.4%~48%。结论基于本文提出的算法融合效果优越,具有较高普适性和实用性,是一种可行的、有效的医学图像融合算法。Objective In this paper,a novel method for the fusion of multimodal medical images is proposed based on nonsubsampled contourlet transform(NSCT).Methods Firstly,the source medical images were initially transformed by NSCT followed by fusing low and high frequency sub-bands.Secondly,the low frequency components of NSCT were fused by the maximum local mean scheme and high frequency components were fused by the maximum local variance scheme.Thirdly,the use of variance enhanced the fusion scheme by preserving the edges in the images.These combinations significantly preserved more details in the source images and improved the quality of the fused images.Finally,the fused image was reconstructed by inverse non-subsampled contourlet transform.Results The efficiency of the suggested technique was carried out by fusion experiments on6different multimodality medical image pairs,visually and quantitatively experimental results indicated that the percentage improvement in entropy,mean,standard deviation and edge strength in proposed method were0~40%,3%~42%,1%~42%,and0.4%~48%compared to conventional methods for6pairs of medical images.Conclusion The proposed method can obtain more efficient and accurate fusion results.It can provide better robustness,superiority and become a feasible image fusion algorithm.

关 键 词：非下采样轮廓波变换 多模态图像融合 融合规则 医学影像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]