摘要
基于高频信息特征融合的超分辨率对抗网络,通过将局部和全局信息进行分离并融合,以达到提高图像的超分辨率的目的,同时避免过度平滑或者失真现象的发生。本次研究中,相关工作人员基于马尔科夫判别器和离散小波变换等概念,实现对于图像高频信息特征的提取与融合,搭建网络整体架构并引入损失函数。实验结果表明,该方法在图像重建、清晰度和结构方面都优于其他基线模型,并且具有较强的泛化能力和鲁棒性。
A super-resolution adversarial network based on high-frequency information feature fusion aims to improve the super-resolution of images by separating and fusing local and global information,while avoiding excessive smoothing or distortion.In this research,relevant staff realized the extraction and fusion of high-frequency information features of images,built the overall network architecture and introduced loss function based on concepts such as Markov discriminator and discrete wavelet transform.The experimental results show that this method outperforms other baseline models in image reconstruction,clarity,and structure,and has strong generalization ability and robustness.
作者
王永红
WANG Yonghong(Zhengzhou Institute of Finance and Technology,Zhengzhou Henan 450000,China)
出处
《信息与电脑》
2023年第11期103-105,共3页
Information & Computer
关键词
高频特征
特征融合
损失函数
high-frequency features
feature fusion
loss function
