基于生成对抗网络的遮挡图像修复算法

doi:10.13190/j.jbupt.2017-247

北京邮电大学学报 ›› 2018, Vol. 41 ›› Issue (3): 81-86.doi: 10.13190/j.jbupt.2017-247

基于生成对抗网络的遮挡图像修复算法

曹志义, 牛少彰, 张继威

北京邮电大学智能通信软件与多媒体北京市重点实验室, 北京 100876

收稿日期:2017-12-04 出版日期:2018-06-28 发布日期:2018-06-04
作者简介:曹志义(1981-),男,博士生;牛少彰(1963-),男,教授,博士生导师,E-mail:szniu@bupt.edu.cn.
基金资助:
国家自然科学基金项目（U1536121，61370195）

Masked Image Inpainting Algorithm Based on Generative Adversarial Nets

CAO Zhi-yi, NIU Shao-zhang, ZHANG Ji-wei

Beijing Key Laboratory of Intelligent Telecommunication Software and Multimedia, Beijing University of Posts and Telecommunications, Beijing 100876, China

Received:2017-12-04 Online:2018-06-28 Published:2018-06-04

摘要/Abstract

摘要： 提出一种基于生成对抗网络的遮挡图像修复算法，能够在大量像素缺失的场景下复原出图像的本来面目.该算法不同于其他的样本块搜索复原算法，可直接生成并且填充可能的缺失元素，改进了生成对抗网络生成模型的结构和生成损失的计算方法，具有半监督学习的特点.实验结果表明，在满足图像整体轮廓的前提下，新算法优于其他算法.

关键词: 生成对抗网络, 半监督学习, 遮挡图像修复

Abstract: A masked image inpainting algorithm based on generative adversarial nets was proposed, which can restore the original image from the lacking of a large number of pixels. Unlike other block search restoration algorithms, the algorithm proposed directly generates possible missing elements and restore them. Due to the improved structure of generated model and the calculation method of generating loss on generative adversarial nets, this article has the characteristics of semi-supervised learning. Experiments show that the proposed method outperforms the existing one on the premise of satisfying the overall contour of the image.

Key words: generative adversarial nets, semisupervised learning, masked image inpainting

中图分类号:

TN929.53

曹志义, 牛少彰, 张继威. 基于生成对抗网络的遮挡图像修复算法[J]. 北京邮电大学学报, 2018, 41(3): 81-86.

CAO Zhi-yi, NIU Shao-zhang, ZHANG Ji-wei. Masked Image Inpainting Algorithm Based on Generative Adversarial Nets[J]. JOURNAL OF BEIJING UNIVERSITY OF POSTS AND TELECOM, 2018, 41(3): 81-86.

参考文献

[1] Bertalmio M, Sapiro G, Caselles V, et al. Image inpainting[C]//Conference on Computer Graphics and Interactive Techniques. New Orleans, LA:ACM Press/Addison-Wesley Publishing Company, 2000:417-424.
[2] Criminisi A, Perez P, Toyama K. Region filling and object removal by exemplar-based image inpainting[J]. IEEE Transactions on Image Processing, 2004, 13(9):1200-1212.
[3] Wong A, Orchard J. A nonlocal-means approach to exemplar-based inpainting[C]//IEEE International Conference on Image Processing. San Antonio:IEEE, 2008:2600-2603.
[4] Bugeau A, Bertalmio M, Caselles V, et al. A comprehensive framework for image inpainting[J]. IEEE Transactions on Image Processing, 2010, 19(10):2634-2645.
[5] Wu J, Ruan Q, An G. Exemplar-based image completion model employing PDE corrections[J]. Informatica, 2008, 21(2):259-276.
[6] Li Y, Liu S, Yang J, et al. Generative face completion[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Hawaii:Convention Center, 2017:6.
[7] Goodfellow I J, Pougetabadie J, Mirza M, et al. Generative adversarial nets[J]. Advances in Neural Information Processing Systems, 2014, 3(2):2672-2680.
[8] Radford A, Metz L, Chintala S. Unsupervised representation learning with deep convolutional generative adversarial networks[J]. CoRR, 2015, 11(06434):1-7.
[9] Arjovsky M, Chintala S, Bottou L. Wasserstein gan[J]. arXiv preprint arXiv:1701. 07875, 2017. https://arxiv.org/abs/1701.07875.
[10] Van den Oord A, Kalchbrenner N, Espeholt L, et al. Conditional image generation with pixelcnn decoders[C]//Advances in Neural Information Processing Systems. Barcelona:[s.n.], 2016:4790-4798.
[11] Kingma D P, Ba J. Adam:a method for stochastic optimization[J]. arXiv preprint arXiv:1412. 6980, 2014. https://arxiv.org/abs/1412.6980.

基于生成对抗网络的遮挡图像修复算法

Masked Image Inpainting Algorithm Based on Generative Adversarial Nets

PDF

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 8

编辑推荐

Metrics

本文评价

[1]	许同乐, 苏元浩, 孟良, 兰孝升, 李云凤. 优化GAN的风电机组齿轮箱故障诊断方法[J]. 北京邮电大学学报, 2023, 46(3): 62-66.
[2]	贾珺, 冯春燕, 夏海轮, 张天魁, 李成钢. 基于样本均衡与特征交互的通信网络故障预测方法[J]. 北京邮电大学学报, 2021, 44(6): 59-66.
[3]	霍纬纲, 王星, 梁锐. 融合互信息估计和对抗自编码器的异常检测[J]. 北京邮电大学学报, 2021, 44(5): 28-34.
[4]	王辰魁, 陈岳林, 蔡晓东. 图像样式风格迁移的行人再识别方法[J]. 北京邮电大学学报, 2021, 44(3): 67-72.
[5]	任晓琪, 翁仲铭. 基于生成对抗网络模型的RSS缺失值预测[J]. 北京邮电大学学报, 2021, 44(1): 26-31.
[6]	王蓬辉, 李明正, 李思. 基于数据增强的中文医疗命名实体识别[J]. 北京邮电大学学报, 2020, 43(5): 84-90.
[7]	刘珍, 王若愚, 刘琼. 基于Bootstrapping的因特网流量分类方法[J]. 北京邮电大学学报, 2014, 37(5): 66-70,79.
[8]	张韬政,王小捷,仝辉. 交互结构最优化算法中辅助问题的相关性[J]. 北京邮电大学学报, 2011, 34(4): 14-18.