CNN-based Euler's Elastica Inpainting with Deep Energy and Deep Image Prior

被引:2
|
作者
Schrader, Karl [1 ]
Alt, Tobias [1 ]
Weickert, Joachim [1 ]
Ertel, Michael [1 ]
机构
[1] Saarland Univ, Math Image Anal Grp, Fac Math & Comp Sci, Campus E1-7, D-66041 Saarbrucken, Germany
基金
欧洲研究理事会;
关键词
image inpainting; variational methods; CNNs; deep energy; deep image prior; Euler's elastica;
D O I
10.1109/EUVIP53989.2022.9922788
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
Euler's elastica constitute an appealing variational image inpainting model. It minimises an energy that involves the total variation as well as the level line curvature. These components are transparent and make it attractive for shape completion tasks. However, its gradient flow is a singular, anisotropic, and nonlinear PDE of fourth order, which is numerically challenging: It is difficult to find efficient algorithms that offer sharp edges and good rotation invariance. As a remedy, we design the first neural algorithm that simulates inpainting with Euler's Elastica. We use the deep energy concept which employs the variational energy as neural network loss. Furthermore, we pair it with a deep image prior where the network architecture itself acts as a prior. This yields better inpaintings by steering the optimisation trajectory closer to the desired solution. Our results are qualitatively on par with state-of-the-art algorithms on elastica-based shape completion. They combine good rotation invariance with sharp edges. Moreover, we benefit from the high efficiency and effortless parallelisation within a neural framework. Our neural elastica approach only requires 3x3 central difference stencils. It is thus much simpler than other well-performing algorithms for elastica inpainting. Last but not least, it is unsupervised as it requires no ground truth training data.
引用
收藏
页数:6
相关论文
共 50 条
  • [1] CNN-Based Stereoscopic Image Inpainting
    Chen, Shen
    Ma, Wei
    Qin, Yue
    [J]. IMAGE AND GRAPHICS, ICIG 2019, PT III, 2019, 11903 : 95 - 106
  • [2] Euler's elastica and curvature-based inpainting
    Chan, Tony F.
    Kang, Sung Ha
    Shen, Jianhong
    [J]. SIAM Journal on Applied Mathematics, 2002, 63 (02) : 564 - 592
  • [3] Deep CNN-Based Blind Image Quality Predictor
    Kim, Jongyoo
    Anh-Duc Nguyen
    Lee, Sanghoon
    [J]. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS, 2019, 30 (01) : 11 - 24
  • [4] Euler's elastica and curvature-based inpainting
    Chan, TF
    Kang, SH
    Shen, JH
    [J]. SIAM JOURNAL ON APPLIED MATHEMATICS, 2003, 63 (02) : 564 - 592
  • [5] A transformer–CNN for deep image inpainting forensics
    Xinshan Zhu
    Junyan Lu
    Honghao Ren
    Hongquan Wang
    Biao Sun
    [J]. The Visual Computer, 2023, 39 : 4721 - 4735
  • [6] Mixture of Deep CNN-based Fnsemble Model for Image Retrieval
    Huang, Hsin-Kai
    Chiu, Chien-Fang
    Kuo, Chien-Hao
    Wu, Yu-Chi
    Chu, Narisa N. Y.
    Chang, Pao-Chi
    [J]. 2016 IEEE 5TH GLOBAL CONFERENCE ON CONSUMER ELECTRONICS, 2016,
  • [7] Deep Structured Energy-Based Image Inpainting
    Altinel, Fazil
    Ozay, Mete
    Okatani, Takayuki
    [J]. 2018 24TH INTERNATIONAL CONFERENCE ON PATTERN RECOGNITION (ICPR), 2018, : 423 - 428
  • [8] A transformer-CNN for deep image inpainting forensics
    Zhu, Xinshan
    Lu, Junyan
    Ren, Honghao
    Wang, Hongquan
    Sun, Biao
    [J]. VISUAL COMPUTER, 2023, 39 (10): : 4721 - 4735
  • [9] IMAGE RESTORATION OF MEDICAL IMAGES WITH STREAKING ARTIFACTS BY EULER'S ELASTICA INPAINTING
    Zhang, Xiaochen
    Wan, Justin W. L.
    [J]. 2017 IEEE 14TH INTERNATIONAL SYMPOSIUM ON BIOMEDICAL IMAGING (ISBI 2017), 2017, : 235 - 239
  • [10] Adaptive Coati Optimization Enabled Deep CNN-based Image Captioning
    Balasubramaniam, S.
    Kadry, Seifedine
    Dhanaraj, Rajesh Kumar
    Kumar, K. Satheesh
    [J]. APPLIED ARTIFICIAL INTELLIGENCE, 2024, 38 (01)