基于区域像素差绝对值总和的NSST-PCNN医学图像融合  

Medical image fusion algorithm with NSST-PCNN based on the sum of absolute value of regional pixel difference

作　　者：黄陈建 戴文战[1] HUANG Chenjian;DAI Wenzhan(School of Information and Electronic Engineering,Zhejiang Gongshang University,Hangzhou,Zhejiang 310018,China)

机构地区：[1]浙江工商大学信息与电子工程学院,浙江杭州310018

出　　处：《光电子．激光》2021年第6期587-594,共8页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金资助项目(61374022)资助项目。

摘　　要：在诸多医学图像融合方法中,非下采样剪切波(non-down sampling shear wave transporm,NSST)与脉冲耦合神经网络(pulse coupled neural network,PCNN)具有较大优势。提出一种基于区域像素差绝对值总和的NSST与PCNN的医学图像融合算法。该算法先将两幅源图像采用NSST进行分解获得低频子带系数与高频子带系数;再将两幅源图像的低频子带系数采用基于区域像素差绝对值总和的规则进行融合;高频子带系数采用基于区域像素差绝对值总和对PCNN的参数进行设置,再利用PCNN获取融合图像高频系数;最后,经过逆NSST获得融合图像。大量实验证明,提出的融合算法较目前其他主流算法具有明显的优势,能同时保留源图像能量和细节,具有较高的视觉效果。Comprehensive diagnosis which combines information of different medical images has become the current mainstream.Among many methods of medical image fusion,non-down sampling shear wave transform(NSST)and pulse coupled neural network(PCNN)have their advantages.This paper adopts the algorithm for medical image fusion which combines NSST with PCNN.In this method,the two source images are first decomposed by using NSST to obtain low-frequency sub-band coefficients and high-frequency sub-band coefficients;then the low-frequency sub-band coefficients of the two source images are fused by using a rule based on the sum of absolute values of regional pixel differences.For image-fused,the parameters of PCNN are set based on the sum of absolute values of regional pixel differences,and then PCNN is used to obtain high-frequency coefficients of the image-used;finally,the image-used is obtained through inverse NSST.Experiments show that the fusion algorithm proposed in this paper has obvious advantages over other algorithms at present.It retains the energy and details of the source image and has a high visual effect.

关 键 词：图像融合 区域像素差绝对值总和 客观评价指标 

分 类 号：TP391.4.1[自动化与计算机技术—计算机应用技术]