文档详情

基于全卷积神经网络的图像修复算法研究.pdf

发布:2021-10-21约3.69万字共36页下载文档
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
相似文档