结合稀疏理论与非下采样剪切波变换的多模态医学图像融合算法  

Multi-modality medical image fusion algorithm combined with sparse theory and non-subsampled shearlet transform

作　　者：李晓军 戴文战[2] 李俊峰[1] LI Xiaojun;DAI Wenzhan;LI Junfeng(Faculty of Mechanical Engineering ＆ Automation,Zhejiang Sci Tech University,Hangzhou 310018,China;School of Information and Electronic,Zhejiang Gongshang University,Hangzhou 310018,China)

机构地区：[1]浙江理工大学机械与自动控制学院,杭州310018 [2]浙江工商大学信息与电子工程学院,杭州310018

出　　处：《浙江理工大学学报（自然科学版）》2018年第6期723-731,共9页Journal of Zhejiang Sci-Tech University(Natural Sciences)

基　　金：国家自然科学基金项目(61374022)

摘　　要：针对医学图像时常存在局部信息丢失、细节模糊不清等问题,为提高可视化效果,避免医疗误诊,提出了一种新的非下采样剪切波变换(NSST)算法。首先,利用NSST分解源图像得到高频子带系数与低频子带系数;其次,根据高低频子带系数的不同特性制定不同的融合策略,对稀疏性不佳的低频系数采用稀疏理论进行处理,反映图像细节信息的高频子带通过相对标准差比较法(RSDC)进行处理;最后,将融合后的高低频子带系数通过NSST重构得到最终的融合图像。从定性和定量角度来评价融合后的图像,算法融合效果良好,与其它5种算法比较发现:该算法在标准差、边缘信息评价因子等指标上表现较好,其余指标均处于中上水平。实验结果表明,该算法得到的融合图像在信息丰富性、对比度、清晰度等方面表现较好,有效增加了不同模态之间的互补信息,具有较好的应用前景。Some problems occur in medical images, such as the loss of local information, and the blurring details. To improve the visual effect and avoid medical misdiagnosis, a new NSST algorithm is proposed in this paper. Firstly, the coefficient of the high frequency sub band and the low frequency sub band coefficient were obtained by using the NSST to decompose source image. Secondly, according to the different characteristics of the high and low frequency sub band coefficients, different fusion strategies were formulated. The low frequency coefficient with poor sparsity was treated with sparse theory. Moreover, the high frequency sub bands that reflect the details of the image were processed by the relative standard deviation comparison （RSDC） method. Finally, the fused high and low frequency sub band coefficients were reconstructed by NSST to gain the final fusion images. The fusion image was evaluated from qualitative and quantitative perspectives, and the algorithm in this paper has a good fusion effect. Compared with the other 5 algorithms, the algorithm in this paper is especially prominent in terms of standard deviation, marginal information evaluation factor and other indicators, and the remaining indicators are in the upper middle levels. The experiment indicates that the fusion image obtained by this algorithm is superior in information richness, contrast, definition and so on. It effectively increases complementary information between different modalities, and has a good application prospect.

关 键 词：图像融合 稀疏理论 NSST变换 相对标准差比较法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]