一种改进的基于NSST-SPCNN医学图像融合算法  

An improved medical image fusion algorithm based on NSST-SPCNN

作　　者：常春红 王雷[1] 郝本利 邢艺馨 CHANG Chunhong;WANG Lei;HAO Benli;XING Yixin(School of Computer Science and Technology,Shandong University of Technology,Zibo 255049,China)

机构地区：[1]山东理工大学计算机科学与技术学院,山东淄博255049

出　　处：《山东理工大学学报（自然科学版）》2021年第4期11-17,共7页Journal of Shandong University of Technology:Natural Science Edition

基　　金：国家自然科学基金项目(61502282);山东省自然科学基金项目(ZR2015FQ005);山东省高等学校科技计划项目(J18KA362)。

摘　　要：为了尽可能多地获取图像中的细节与边缘信息,提出了一种基于非亚采样剪切波变换和改进自适应脉冲耦合神经网络相结合的图像融合算法。采用非亚采样剪切波变换算法将两幅精配准的图像进行分解,分别得到两幅图像的低频分量与不同尺度方向的高频分量。在低频系数区采取局部能量加权和与双边滤波来计算融合不同尺度的低通分量,实现细节的提取与能量的保存。在高频系数区域,采用改进的自适应参数脉冲耦合神经网络算法,通过简化脉冲耦合神经网络模型、优化自适应参数融合高通分量,提高融合的效率与质量,同时避免人工输入经验阈值的不便。最后,经过NSST的逆变换得到最终的融合图像。实验结果表明,该算法能有效地保持图像边缘与纹理,保留图像的细节信息与纹理特征。与传统算法相比,具有更好的性能与适用性。In order to obtain as much detail and edge information as possible,an image fusion algorithm based on the non-subsampling shear wave transform and improved adaptive pulse coupled neural network is proposed.In this paper,the non-subsampling shear wave transform algorithm is used to decompose the two precisely registered images,and the low-frequency components of the two images and the high-frequency components of different scales are obtained respectively.In the low-frequency region,local energy weighted sum and bilateral filter are used to calculate and fuse low-pass components of different scales,so as to extract details and save energy.In the region of high-frequency coefficients,the improved adaptive parameter pulse coupled neural network algorithm is used to simplify the pulse coupled neural network model and optimize the adaptive parameter to fuse high-pass components,which improves the efficiency and quality of fusion and avoids the inconvenience of manual input experience threshold.Finally,the final fusion image is obtained by the inverse transform of non-subsampling shear wave transform.The experimental results show that this method can effectively maintain the image edge and texture,and retain the image details and texture features.Compared with the traditional method,it has better performance and applicability.

关 键 词：医学图像融合 脉冲耦合神经网络 自适应参数 非亚采样剪切波变换 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]