结合扩张金字塔的脑部医学图像融合  

Medical Brain Image Fusion with Dilated Pyramid

作　　者：马为民 郑茜颖[1] MA Weimin;ZHENG Qianying(College of Physics&Information Engineering,Fuzhou University,Fuzhou 350108,China)

机构地区：[1]福州大学物理与信息工程学院,福建福州350108

出　　处：《电视技术》2024年第1期16-21,29,共7页Video Engineering

基　　金：国家自然科学基金(62271151);福建省科技重点产业引导项目(2020H0007)。

摘　　要：针对现有脑部医学图像融合算法存在的融合图像细节模糊和边缘性差等问题,设计一种扩张金字塔特征提取算法,由特征提取器、特征融合器和特征重构器3部分组成。特征提取器由扩张金字塔特征模块提取浅层和深层图像特征的结合,防止图像细节信息的丢失;特征融合器采用改进的功能能量比(Functional Energy Ratio,FER)特征融合策略增强融合图像边缘信息;特征重构器由4层卷积构成归一化图像。实验结果表明,相较于当前通用的脑部融合算法,所提出的算法具有较好的视觉效果和细节信息,客观评价指标有更好的表现。Aiming at the problems of blurred detail and poor borderline i nthe existing brain medical image fusion algorithms,this pape r designs an expanded pyramid feature extraction algorithm,which consists of three parts:feature extractor,feature fusion and feature reconstructor.The feature extractor extracts the combination of shallow and deep image features by the expanding pyramid feautre module to prevent the loss of image details.The improved Functional Energy Ratio(FER)feature fusion strategy is used to enhance the edge information of the fused image.The feature reconfiguarble image is normalized by 4 layers of convolution.The expemri ental results show that compared with the current general brain fusion algorithm,the proposed algorithm has better visual effect and detailed information,and better performance of objective evaluation indicators.

关 键 词：脑部医学图像融合 多模态医学图像 金字塔特征 特征融合 特征重构 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]