摘要
随着数码相机、手机等电子设备的普及,每天都会产生大量的图像,但通常这些图像的分辨率比较低。针对单幅图像超分辨率(Single Image Super-Resolution,SISR)方法性能较低的问题,提出一种基于残差密集网络的单幅图像超分辨率重建方法。将浅层的卷积特征输入到残差密集块,获得全局和局部的特征;对图像进行超分辨率重建,得到清晰的高分辨率图像。为了验证该方法的有效性,在四个公共的数据集Set5、Set14、B100和Urban10上进行了定性和定量的实验。实验结果表明,该方法能够更好地恢复出高分辨率的图像。
With the popularity of electronic devices such as digital cameras and mobile phones,a large number of images are produced every day,but usually the resolution of these images is low.Aiming at the problem of low performance of single image super-resolution(SISR) method,this paper proposed a super-resolution reconstruction method for single image based on residual-dense network.I input the convolution feature of shallow layer into residual-dense blocks to obtain global and local features,and reconstructed the image with super resolution to get clear and high resolution image.To verify the effectiveness of this method,this paper conducted qualitative and quantitative experiments on four public data sets Set5,Set14,B100 and Urban10.Experimental results show that the proposed method can better restore high resolution images.
作者
谢雪晴
Xie Xueqing(College of of Information Engineering,Chongqing Industry Polytechnic College,Chongqing 401120,China)
出处
《计算机应用与软件》
北大核心
2019年第10期222-226,共5页
Computer Applications and Software
基金
重庆市教委科学技术研究项目(KJ1603701)
关键词
图像超分辨率
低分辨率
高分辨率
深度学习
Image super-resolution
Low resolution
High resolution
Deep learning
