基于全卷积神经网络的图像修复算法研究.pdf
文本预览下载声明
基于全卷积神经网络的图像修复算法研究
161203105333
·
2020
的基于全卷积神经网络的图像修
复算法研究的研究的
的
的
2020
基于全卷积神经网络的图像修复算法研究
图像修复算的研究的
图像修复算法基于图像
积的
的图像修复的的基于的图像修复算法
于的修复的
网络的
基于全卷积神经网络的图像修复算法神经网络的
网络图像经全全图像
图像全网络的网络全
的修复的图像图像的的
的图像的图像
CelebA
算法修复
图像修复 全卷积神经网络 神经网络
2020
Researchon image inpainting based onFullyConvolutional
Networks
Abstract
Image inpainting is a difficult problem in the field of deep learning. Traditional image
inpainting algorithms are mostly based on image diffusion or patch, and usually require
smoothness prior. This algorithms often perform poorly in scenes with large missing areas
or strong semantic scenes. The development of deep learning technology has made image
repair have a new research direction. Image inpainting algorithms based on deep learning
can be applied to a wider range of scenes, greatly improving the repair effect. However,
the depth generation model still has great room for improvement, such as single network
structure,unstabletrainingprocess,poorinterpretability,andlackoftheoreticalcertification
forsuperparameterselection.
ThispaperbringforwardanimageinpaintingalgorithmbasedonFCN.Withreferencetothe
ideaofGAN,thenetworkstructureofthecontextencoderisimproved,sothatthepartially
missing image is visually consistent with the global image after completio
显示全部