摘要
当前图像融合方法存在融合精度低的问题,融合图像清晰度不够,质量低,为了解决当前图像融合方法存在的这些问题,获得更高质量的图像融合结果,提出了基于卷积神经网络的图像融合方法。首先采集待融合图像,并对它们进行预处理,然后分别将预处理后的图像输入到卷积神经网络进行训练,提取它们的图像融合特征,最后采用最优阈值法对融合特征进行分割,对不同图像不同区域进行相应的融合,得到最终的图像融合结果,并采用具体图像融合实验分析了卷积神经网络和其它方法的性能,结果表明,卷积神经网络融合后的图像不仅清晰度和亮度得到了明显的改善,而且提高了图像信噪比,图像质量更高,具有十分明显的优越性。
The current image fusion method has the problem of low fusion accuracy,which makes the fusion image not clear enough and low quality.In order to solve the problem and obtain higher quality image fusion result,an image fusion method based on convolution neural network is proposed.Firstly,the images to be fused are collected and preprocessed,and then the processed images are input to convolution neural network for training,and their image fusion features are extracted.Finally,the optimal threshold method is used to segment the fusion features,and different regions of different images are fused accordingly,and the final image fusion results are obtained.The performances of convolution neural network and other methods are analyzed.The results show that the image clarity and brightness are improved obviously after the convolution neural network fusion,and the image signal-to-noise ratio is improved,the image quality is higher,which has obvious advantages.
作者
李婧
保慧琴
李茹
LI Jing;BAO Huiqin;LI Ru(School of Information Engineering,Xi’an Mingde Institute of Technology,Xi’an 710124,China)
出处
《微型电脑应用》
2021年第8期32-34,38,共4页
Microcomputer Applications
基金
陕西省教育厅专项科研计划项目(20JK0954)
西安明德理工学院2019年科研基金项目(2019JK0954)。
关键词
图像质量
融合方法
卷积神经网络
采样剪切波变换
不同频率子带
image quality
fusion method
convolutional neural network
sampled shear wave transform
different frequency sub-bands
