摘要
为提升融合图像清晰度,提出一种基于引导滤波器与自适应稀疏表示的多模态医学图像融合算法。该算法利用高斯滤波器将输入图像分解为细节层和基础层;基于显著性特征和引导滤波器求得基础层权值图,根据该权值图结合加权平均融合规则对基础层进行融合;同时,采用自适应稀疏表示算法融合细节层;最后,将融合的细节层和基础层相加得到融合图像。在指标评价和视觉分析上比较了该算法和其他6种经典算法的融合结果;此外,还比较了该算法与两种基于稀疏表示算法的时间复杂度。结果表明,该算法在纹理和边缘信息保存上优于其他算法,其时间复杂度优于基于稀疏表示的算法。
In order to improve the clarity of fused images,a multi-modal medical image fusion algorithm based on guided filter and adaptive sparse representation is proposed.Specifically,this algorithm adopts Gaussian filter to decompose the input images into detail layers and base layers.Subsequently,the weight maps of base layers are obtained based on saliency characteristics and guided filters,which are further utilized to fuse the base layers in combination with the weighted average rule;at the same time,the detail layers are fused through an adaptive sparse representation algorithm.Finally,the fused layers are directly added into the base layers to obtain the fused image.This algorithm is compared with other six classical algorithms on quality evaluation and visual analysis.In addition,the time complexity of the algorithm is also compared with that of two sparse representation-based algorithms.The results show that this algorithm outperforms other algorithms in the preservation of texture and edge information.Meanwhile,its time complexity is significantly better than that of sparse representation-based algorithms.
作者
王兆滨
马一鲲
崔子婧
WANG Zhaobin;MA Yikun;CUI Zijing(School of Information Science and Engineering,Lanzhou University,Lanzhou 730000)
出处
《电子科技大学学报》
EI
CAS
CSCD
北大核心
2022年第2期264-273,共10页
Journal of University of Electronic Science and Technology of China
基金
国家自然科学基金(61201421)。
