基于监督学习深度自编码器的图像重构

doi:10.11896／j.issn.1002-137X.2018.11.042

计算机科学 ›› 2018, Vol. 45 ›› Issue (11): 267-271.doi: 10.11896／j.issn.1002-137X.2018.11.042

基于监督学习深度自编码器的图像重构

张赛¹, 芮挺^1,2, 任桐炜², 杨成松¹, 邹军华¹

(陆军工程大学野战工程学院南京210007)¹
(南京大学计算机软件新技术国家重点实验室南京210023)²

收稿日期:2017-10-11 发布日期:2019-02-25
作者简介:张赛(1991-),男,硕士生,主要研究方向为机器学习、计算机视觉,E-mail:466908114@qq.com;芮挺(1972-),男,博士,副教授,CCF高级会员,主要研究方向为机器学习、计算机视觉,E-mail:rtinguu@sohu.com(通信作者);任桐炜(1981-),男,博士,副教授,主要研究方向为视觉媒体计算;杨成松(1982-),男,博士,讲师,主要研究方向为信息检索、数字图像处理;邹军华(1991-),男,博士生,主要研究方向为机器学习、计算机视觉。
基金资助:
本文受国家自然科学基金:数据驱动的社会媒体信息传播演化关键技术研究(61473444)资助。

Image Reconstruction Based on Supervised Learning Deep Auto-encoder

ZHANG Sai¹, RUI Ting^1,2, REN Tong-wei², YANG Cheng-song¹, ZOU Jun-hua¹

(Department of Filed Engineering,PLA Army Engineering University,Nanjing 210007,China)¹
(State Key Laboratory for Novel Software Technology,Nanjing University,Nanjing 210023,China)²

Received:2017-10-11 Published:2019-02-25

摘要/Abstract

摘要： 针对数字图像受损信息的重构问题,提出一种将经典无监督学习自编码器(Auto-Encoder,AE)用于监督学习的新方法,并对深度模型结构与训练策略进行了研究。通过设计多组监督学习单层AE模型,提出了逐组“递进学习”和“关联编码”的学习策略,构建了一个新的基于监督学习的深度AE模型结构;对于新模型结构,采用多对一(一个输入样本的多种形式对应一个输出)的训练方法代替经典AE中一对一(一个输入样本对应一个输出)的训练方法。将该模型的结构和训练策略用于部分数据受损或遮挡的图像中进行数据重构测试,提高了模型对受损数据特征编码的表达能力和重构能力。实验结果表明,提出的新方法对于受损及遮挡样本的图像具有良好的重构效果和适应性。

关键词: 监督学习, 深度结构, 图像重构, 训练策略, 自编码器

Abstract: Aiming at the reconstruction of the damaged information of digital image,this paper proposed a new approach in which the classical unsupervised auto-encoder(AE) is used for supervised learning,and researched the deep model structure and training strategy.Specifically,this paper presented a novel supervised learning based deep auto-encoder model which possesses a set of progressive and interrelated learning strategies through designing multiple groups of supervised single-layer AE.In the novel model,the one-to-one training strategy in classical AE model (one output corresponding to one input) is substituted by the many-to-one training strategy (one output corresponding to many inputs).Then,the structure and training strategy mentioned above were utilized for the damaged or occluded images to test the process of data reconstruction,thus improving the model’s ability to express and reconstruct the feature of those data.Experimental results show that the new method has good reconstruction effect and adaptability to the damaged or occluded samples.

Key words: Auto-encoder, Deep structure, Image reconstruction, Supervised learning, Training strategy

中图分类号:

TP391

张赛, 芮挺, 任桐炜, 杨成松, 邹军华. 基于监督学习深度自编码器的图像重构[J]. 计算机科学, 2018, 45(11): 267-271. https://doi.org/10.11896／j.issn.1002-137X.2018.11.042

ZHANG Sai, RUI Ting, REN Tong-wei, YANG Cheng-song, ZOU Jun-hua. Image Reconstruction Based on Supervised Learning Deep Auto-encoder[J]. Computer Science, 2018, 45(11): 267-271. https://doi.org/10.11896／j.issn.1002-137X.2018.11.042

参考文献

[1]ZHANG H Y,PENG Q C.A survey on digital image inpainting[J].Journal of Image & Graphics,2007,12(1):1-10.(in Chinese)
张红英,彭启琮.数字图像修复技术综述[J].中国图象图形学报,2007,12(1):1-10.
[2]BERTALMÍO M,SAPIRO G,CASELLES V,et al.Image in- painting[C]∥Conference on Computer Graphics and Interactive Techniques.DBLP,2000:417-424.
[3]CHAN T,SHEN J.Mathematical models for local non-texture inpaintings[J].SIAM Journal on Applied Mathematics,2001,62(3):1019-1043.
[4]KOMODAKIS N.Image Completion Using Global Optimization[C]∥2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.IEEE,2006:442-452.
[5]HSIN H F,LEOU J J,LIN C S,et al.Image inpaintingusingstructure-guided priority belief propagation and label transformations[C]∥International Conference on Pattern Recognition.IEEE,2010:4492-4495.
[6]MANSFIELD A,PRASAD M,ROTHER C,et al.Transforming Image Completion[C]∥British Machine Vision Conference.2011.
[7]KANG Y,LEE K T,EUN J,et al.Stacked Denoising Autoencoders for Face Pose Normalization[C]∥International Con-ference on Neural Information Processing.Springer Berlin Heidelberg,2013:241-248.
[8]XIE J,XU L,CHEN E.Image denoising and inpainting with deep neural networks[C]∥International Conference on Neural Information Processing Systems.2012:341-349.
[9]PATHAK D,KRAHENBUHL P,DONAHUE J,et al.Context Encoders:Feature learning by inpainting[C]∥2016 IEEE Conference on Computer Vision and Pattern Recognition(CVPR).2016:2536-2544.
[10]YANG C,LU X,LIN Z,et al.High-Resolution Image Inpainting Using Multi-scale Neural Patch Synthesis[C]∥IEEECon-ference on Computer Vision and Pattern Recognition.IEEE,2017:4076-4084.
[11]HUANG R,CHANG L,LIC G,et al.Adaptive Deep Supervised Autoencoder Based Image Reconstruction for Face Recognition[J].Mathematical Problems in Engineering,2016,2016(5):1-14.
[12]HINTON G E,OSINDERO S,TEH Y W.A fast learning algorithm for deep belief nets[J].Neural Computation,2006,18(7):1527.
[13]BENGIO Y.Learning deep architectures for AI[J].Foundations and Trends in Machine Learning,2009,2(1):1-127.
[14]ZHANG C,ZHANG Z Y.Improving multiview face detection with multi-task deep convolutional neural networks[C]∥IEEE Winter Application and Computer Vision Conference.USA,2014:1036-1041.
[15]LÄNGKVIST M,KARLSSON L,LOUTFI A.A review of unsupervised feature learning and deep learning for time-series modeling[J].Pattern Recognition Letters,2014,42(1):11-24.
[16]ZHANG Y,CHEN Q Y,ZHANG Y J.Deep learning and its new progress in object and behavior recognition[J].Journal of Image & Graphics,2014,19(2):175-184.(in Chinese)
郑胤,陈权崎,章毓晋.深度学习及其在目标和行为识别中的新进展[J].中国图象图形学报,2014,19(2):175-184.
[17]RUMELHART D E,HINTON G E,WILLIAMS R J.Learning representations by back-propagating errors[J].Nature,1986,323:533-536.
[18]PEARSON K.Mathematical contributions to the theory of evolution (III):Regression,heredity,and panmixia[J].Prceedings of the Royal Society of London,1998,187(4):253-318.

相关文章 15

[1]	王冠宇, 钟婷, 冯宇, 周帆. 基于矢量量化编码的协同过滤推荐方法 Collaborative Filtering Recommendation Method Based on Vector Quantization Coding 计算机科学, 2022, 49(9): 48-54. https://doi.org/10.11896/jsjkx.210700109
[2]	宋杰, 梁美玉, 薛哲, 杜军平, 寇菲菲. 基于无监督集群级的科技论文异质图节点表示学习方法 Scientific Paper Heterogeneous Graph Node Representation Learning Method Based onUnsupervised Clustering Level 计算机科学, 2022, 49(9): 64-69. https://doi.org/10.11896/jsjkx.220500196
[3]	武红鑫, 韩萌, 陈志强, 张喜龙, 李慕航. 监督和半监督学习下的多标签分类综述 Survey of Multi-label Classification Based on Supervised and Semi-supervised Learning 计算机科学, 2022, 49(8): 12-25. https://doi.org/10.11896/jsjkx.210700111
[4]	杜航原, 李铎, 王文剑. 一种面向电商网络的异常用户检测方法 Method for Abnormal Users Detection Oriented to E-commerce Network 计算机科学, 2022, 49(7): 170-178. https://doi.org/10.11896/jsjkx.210600092
[5]	胡艳羽, 赵龙, 董祥军. 一种用于癌症分类的两阶段深度特征选择提取算法 Two-stage Deep Feature Selection Extraction Algorithm for Cancer Classification 计算机科学, 2022, 49(7): 73-78. https://doi.org/10.11896/jsjkx.210500092
[6]	郁舒昊, 周辉, 叶春杨, 王太正. SDFA:基于多特征融合的船舶轨迹聚类方法研究 SDFA:Study on Ship Trajectory Clustering Method Based on Multi-feature Fusion 计算机科学, 2022, 49(6A): 256-260. https://doi.org/10.11896/jsjkx.211100253
[7]	庞兴龙, 朱国胜. 基于半监督学习的网络流量分析研究 Survey of Network Traffic Analysis Based on Semi Supervised Learning 计算机科学, 2022, 49(6A): 544-554. https://doi.org/10.11896/jsjkx.210600131
[8]	侯夏晔, 陈海燕, 张兵, 袁立罡, 贾亦真. 一种基于支持向量机的主动度量学习算法 Active Metric Learning Based on Support Vector Machines 计算机科学, 2022, 49(6A): 113-118. https://doi.org/10.11896/jsjkx.210500034
[9]	王宇飞, 陈文. 基于DECORATE集成学习与置信度评估的Tri-training算法 Tri-training Algorithm Based on DECORATE Ensemble Learning and Credibility Assessment 计算机科学, 2022, 49(6): 127-133. https://doi.org/10.11896/jsjkx.211100043
[10]	张文轩, 吴秦. 基于多分支注意力增强的细粒度图像分类 Fine-grained Image Classification Based on Multi-branch Attention-augmentation 计算机科学, 2022, 49(5): 105-112. https://doi.org/10.11896/jsjkx.210100108
[11]	韩洁, 陈俊芬, 李艳, 湛泽聪. 基于自注意力的自监督深度聚类算法 Self-supervised Deep Clustering Algorithm Based on Self-attention 计算机科学, 2022, 49(3): 134-143. https://doi.org/10.11896/jsjkx.210100001
[12]	武玉坤, 李伟, 倪敏雅, 许志骋. 单类支持向量机融合深度自编码器的异常检测模型 Anomaly Detection Model Based on One-class Support Vector Machine Fused Deep Auto-encoder 计算机科学, 2022, 49(3): 144-151. https://doi.org/10.11896/jsjkx.210100142
[13]	许华杰, 陈育, 杨洋, 秦远卓. 基于混合样本自动数据增强技术的半监督学习方法 Semi-supervised Learning Method Based on Automated Mixed Sample Data Augmentation Techniques 计算机科学, 2022, 49(3): 288-293. https://doi.org/10.11896/jsjkx.210100156
[14]	唐雨潇, 王斌君. 基于深度生成模型的人脸编辑研究进展 Research Progress of Face Editing Based on Deep Generative Model 计算机科学, 2022, 49(2): 51-61. https://doi.org/10.11896/jsjkx.210400108
[15]	侯宏旭, 孙硕, 乌尼尔. 蒙汉神经机器翻译研究综述 Survey of Mongolian-Chinese Neural Machine Translation 计算机科学, 2022, 49(1): 31-40. https://doi.org/10.11896/jsjkx.210900006

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

基于监督学习深度自编码器的图像重构

Image Reconstruction Based on Supervised Learning Deep Auto-encoder

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

Metrics

本文评价

推荐阅读 0