摘要
图像融合过程中存在信息丢失的问题,导致融合后的图像质量较低。为此,提出基于小波变换的图像多尺度空间融合算法。将源图像矢量化信息转化为稀疏形式,并通过投影将其划分为若干个大小相等的单元块,在小波多尺度变换的支持下,将单元块转换为小波域,实现图像非均匀分块压缩信息的完整感知,利用滑动窗口将小波域分割为不同的尺度,按照尺度域的梯度幅值、区域能量以及匹配度实现图像的融合。结果表明,融合图像的互信息、边缘保持度、相似度、信息熵、平均梯度分别为7.114、0.877、0.982、7.567、9.436,均明显优于对比方法。
The problem of information loss in the process of image fusion leads to poor quality of the fused image.Therefore,an image multi-scale spatial fusion algorithm based on wavelet transform is proposed.The vectorized information of the source image is transformed into sparse form,and it is divided into several unit blocks of equal size by projection.With the support of wavelet multi-scale transformation,the unit blocks are converted into wavelet domain to realize the complete perception of the compressed information of non-uniform image blocks.The wavelet domain is divided into different scales by sliding window,and the image is fused according to the gradient amplitude,regional energy,and matching degree of the scale domain.The results show that the mutual information,edge retention,similarity,information entropy,and average gradient of the fused image are 7.114,0.877,0.982,7.567,and 9.436,respectively,which are significantly superior to the contrast method.
作者
单盛
SHAN Sheng(School of Software Engineering,Anhui Vocational College of Electronics and Information Technology,Bengbu 233000,Anhui,China)
出处
《上海电机学院学报》
2022年第2期95-99,共5页
Journal of Shanghai Dianji University
基金
安徽省2018年高校人文社会科学研究重点资助项目(SK2018A0838)
安徽省特色专业教学资源库资助项目(2019zyk31)。
关键词
图像处理
多尺度空间
融合算法
小波多尺度变换
压缩信息
梯度幅值
image processing
multi-scale space
fusion algorithm
wavelet multi-scale transform
compress information
gradient amplitude
