Academic Journal of Computing & Information Science, 2020, 3(4); doi: 10.25236/AJCIS.2020.030412.
Xinyan Zhang1, *, Xiaofeng Wang2
1 College of Information Engineering, Shanghai Maritime University, Shanghai 201306, China
2 Shanghai Maritime University, Shanghai 201306, Shanghai, China
*Corresponding author
Generative Adversarial Networks(GANs) is a generation model that learns data distribution through the mutual against between generator network and discriminantor network.It has the advantage of generating clear and sharp samples, and has made progress in the application of image super-resolution and image repair. DeblurGAN solves the problem of end-to-end image deblurring by using conditional Generative Adversarial Networks(cGANs).In order to obtain better deblurring effect, this paper proposes an improvement based on DeblurGAN.Firstly, the method in this paper uses DenseBlock to replace the ResBlock in DeblurGAN, and adds two skip-connections. Finally, depthwise separable convolution is used to replace the common convolution block in the network, so as to reduce the network model, reduce parameters and accelerate the convergence speed of the network. The loss function uses the perceptual loss to ensure content consistency between the generated image and the clear image.
motion blur, dense-connected-convolutional-network(DenseNet), skip-connection, depthwise separable convolution, perceptual loss
Xinyan Zhang, Xiaofeng Wang. Motion deblurring method based on Improved DeblurGAN. Academic Journal of Computing & Information Science (2020), Vol. 3, Issue 4: 102-109. https://doi.org/10.25236/AJCIS.2020.030412.
[1] Helstrom C W. Image Restoration by the Method of Least Squares [J]. J. opt. soc. amrer, 1967, 57 (3): 297-303.
[2] Fergus R, Singh B, Hertzmann A, et al. Removing Camera Shake from a Single Photograph [C] //ACM transactions on graphics (TOG). ACM, 2006, 25 (3): 787-794.
[3] Krishnan D, Fergus R. Fast Image Deconvolution using Hyper-Laplacian Priors [C] //Advances in neural information processing systems. 2009: 1033-1041.
[4] Chakrabarti. A neural approach to blind motion deblurring. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2016.
[5] I. J. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio. Generative Adversarial Networks. June 2014.
[6] S. Nah, T. Hyun, K. Kyoung, and M. Lee. Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring. 2016.
[7] P. Isola, J.-Y. Zhu, T. Zhou, and A. A. Efros. Image-to-image translation with conditional adversarial networks. arxiv, 2016.
[8] KUPYN O, BUDZAN V, MYKHAILYCH M, et al DeblurGAN: blind motion deblurring using conditional adversarial networks [EB/OL].(2018-04-03) [2018-08-15]. https: //arxiv. org /pdf / 1711. 07064.pdf.
[9] Arjovsky M, Chintala S, Bottou L.Wasserstein Generative Adversarial Networks [C]// International Conference on Machine Learning, 2017: 214-223.
[10] Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of Wasserstein GANs [C]//Advances in Neural Information Processing Systems.2017: 5767-5777.
[10] Gulrajani I, Ahmed F, Arjovsky M, et al. Improved training of Wasserstein GANs [C]//Advances in Neural Information Processing Systems.2017: 5767-5777.
[11] SIMONYAN K, ZISSERMAN A. Very deep convolu- tional networks for large-scale image recognition [EB/OL]. (2015-04-10) [2018-08-17]. https://arxiv. org / pdf / 1409. 1556. pdf.
[12] Seungjun Nah, Tae Hyun Kim, and Kyoung Mu Lee. Deep multi-scale convolutional neural network for dynamic scene deblurring. In CVPR, 2017.