Mosaic Removal Algorithm Based on Improved Generative Adversarial Networks Model

  • He Wang
  • Zhiyi Cao
  • Shaozhang NiuEmail author
  • Hui Tong
Conference paper
Part of the Smart Innovation, Systems and Technologies book series (SIST, volume 157)


Generative adversarial networks have yielded outstanding results in unsupervised areas of learning, but existing research has proven that the results are not stable in specific areas. In this paper, an improved generative adversarial networks model is proposed. First, the loss calculation method of the generated model is changed, which makes the removal target of the whole network controllable. Second, the deep convolution network is added to the existing network; this improves the accuracy of the mosaic removal. And then combines the loss calculation method of the pixel networks, the network effectively solve the unstable features of generative adversarial networks in specific conditions. Finally, the experimental results show that the overall mosaic face removal for this network performance is superior to other existing algorithms.


Generative adversarial networks Unsupervised learning Mosaic removal 



This work was supported by National Natural Science Foundation of China (No. U1536121, 61370195).


  1. 1.
    Chung, K.H., Chan, Y.H.: Color demosaicing using variance of color differences. IEEE Trans. Image Process. 15(10), 2944–2955 (2006)CrossRefGoogle Scholar
  2. 2.
    Li, X.: Demosaicing by successive approximation. IEEE Trans. Image Process. A Publ. IEEE Signal Process. Soc. 14(3), 370–379 (2005)Google Scholar
  3. 3.
    Zhang, L., Wu, X.: Color demosaicking via directional linear minimum mean square-error estimation. IEEE Press (2005)Google Scholar
  4. 4.
    Chen, X., Peng, X., Li, J.-B., Peng, Yu.: Overview of deep kernel learning based techniques and applications. J. Netw. Intell. 1(3), 83–98 (2016)Google Scholar
  5. 5.
    Xia, Y., Rong, H.: Fuzzy neural network based energy efficiencies control in the heating energy supply system responding to the changes of user demands. J. Netw. Intell. 2(2), 186–194 (2017)Google Scholar
  6. 6.
    Goodfellow, I.J., Pougetabadie, J., Mirza, M., et al.: Generative adversarial nets. Adv. Neural. Inf. Process. Syst. 3, 2672–2680 (2014)Google Scholar
  7. 7.
    Radford, A., Metz, L., Chintala, S.: Unsupervised representation learning with deep convolutional generative adversarial networks. Comput. Sci. (2015)Google Scholar
  8. 8.
    Arjovsky, M., Chintala, S., Bottou, L.: Wasserstein GAN (2017). arXiv:1701.07875
  9. 9.
    Gulrajani, I., Ahmed, F., Arjovsky, M., et al.: Improved training of Wasserstein GANs (2017). arXiv:1704.00028
  10. 10.
    Oord, A., Kalchbrenner, N., Vinyals, O., et al.: Conditional image generation with PixelCNN decoders (2016). arXiv:1606.05328
  11. 11.
    Li, Y., Liu, S., Yang, J., et al.: Generative face completion (2017). arXiv:1704.05838
  12. 12.
    Kingma, D.P., Ba, J., Lei: Adam: a method for stochastic optimization (2014). arXiv:1412.6980
  13. 13.
    Tieleman, T., Hinton, G.: Lecture 6.5—RmsProp: divide the gradient by a running average of its recent magnitude. In: COURSERA: Neural Networks for Machine Learning (2012)Google Scholar
  14. 14.
    Rasmus, A., Valpola, H., Honkala, M., Berglund, M., Raiko, T.: Semisupervised learning with ladder network (2015). arXiv:1507.02672
  15. 15.
    Ioffe, S., Szegedy, C.: Batch normalization: accelerating deep network training by reducing internal covariate shift (2015). arXiv:1502.03167
  16. 16.
    Dosovitskiy, A., Fischer, P., Springenberg, J.T., Riedmiller, M., Brox, T.: Discriminative unsupervised feature learning with exemplar convolutional neural networks. IEEE Trans. Pattern Anal. Mach. Intell. 99 (2015)Google Scholar
  17. 17.
    Dahl, R., Norouzi, M., Shlens, J.: Pixel recursive super resolution super resolution (2017). arXiv:1702.00783
  18. 18.
    Arjovsky, M., Bottou, L.: Towards principled methods for training generative adversarial networks. NIPS 2016 Workshop on Adversarial TrainingGoogle Scholar

Copyright information

© Springer Nature Singapore Pte Ltd. 2020

Authors and Affiliations

  1. 1.Beijing Key Lab of Intelligent Telecommunication Software and Multimedia, School of Computer ScienceBeijing University of Posts and TelecommunicationsBeijingChina

Personalised recommendations