Lightweight Single Image Super-Resolution by Channel Split Residual Convolution

다국어 초록 (Multilingual Abstract)

In recent years, deep convolutional neural networks have made significant progress in the research of singleimage super-resolution. However, it is difficult to be applied in practical computing terminals or embeddeddevices due to a large number of parameters and computational effort. To balance these problems, we proposeCSRNet, a lightweight neural network based on channel split residual learning structure, to reconstruct highresolutionimages from low-resolution images. Lightweight refers to designing a neural network with fewerparameters and a simplified structure for lower memory consumption and faster inference speed. At the sametime, it is ensured that the performance of recovering high-resolution images is not degraded. In CSRNet, wereduce the parameters and computation by channel split residual learning. Simultaneously, we propose adouble-upsampling network structure to improve the performance of the lightweight super-resolution networkand make it easy to train. Finally, we propose a new evaluation metric for the lightweight approaches named100_FPS. Experiments show that our proposed CSRNet not only speeds up the inference of the neural networkand reduces memory consumption, but also performs well on single image super-resolution.

참고문헌 (Reference)

1 J. B. Huang, "Single image super-resolution from transformed self-exemplars" 5197-5206, 2015

2 X. Zhang, "Shufflenet : an extremely efficient convolutional neural network for mobile devices" 6848-6856, 2018

3 Y. Zhang, "Residual dense network for image super-resolution" 2472-2481, 2018

4 W. Shi, "Real-time single image and video super-resolution using an efficient sub-pixel convolutional neural network" 1874-1883, 2016

5 C. Ledig, "Photo-realistic single image super-resolution using a generative adversarial network" 105-114, 2017

6 E. Agustsson, "Ntire 2017 challenge on single image super-resolution : dataset and study" 1122-1131, 2017

7 X. Chu, "Multi-objective reinforced evolution in mobile neural architecture search"

8 A. G. Howard, "Mobilenets: efficient convolutional neural networks for mobile vision applications"

9 R. Lan, "MADNet : a fast and lightweight network for singleimage super resolution" 51 (51): 1443-1453, 2020

10 M. Bevilacqua, "Low-complexity single-image superresolution based on nonnegative neighbor embedding" 1-10, 2012

11 B. Zoph, "Learning transferable architectures for scalable image recognition" 8697-8710, 2018

12 Y. Tai, "Image super-resolution via deep recursive residual network" 2709-2798, 2017

13 Y. Zhang, "Image super-resolution using very deep residual channel attention networks" 294-310, 2018

14 Z. Wang, "Image quality assessment : from error visibility to structural similarity" 13 (13): 600-612, 2004

15 K. Han, "Ghostnet : more features from cheap operations" 1577-1586, 2020

16 Z. Li, "Feedback network for image super-resolution" 3867-3876, 2019

17 Z. Li, "Feedback network for image super-resolution" 3867-3876, 2019

18 N. Ahn, "Fast, accurate, and lightweight super-resolution with cascading residual network" 256-272, 2018

19 N. Ahn, "Fast, accurate, and lightweight super-resolution with cascading residual network" 256-272, 2018

20 X. Chu, "Fast, accurate and lightweight super-resolution with neural architecture search"

21 Z. Hui, "Fast and accurate single image super-resolution via information distillation network" 723-731, 2018

22 B. Lim, "Enhanced deep residual networks for single image superresolution" 1132-1140, 2017

23 B. Lim, "Enhanced deep residual networks for single image superresolution" 1132-1140, 2017

24 J. Kim, "Deeply-recursive convolutional network for image super-resolution" 1637-1645, 2016

25 K. He, "Deep residual learning for image recognition" 770-778, 2016

26 S. Nah, "Deep multi-scale convolutional neural network for dynamic scene deblurring" 257-265, 2017

27 R. Zeyde, "Curves and Surfaces" Springer 711-730, 2010

28 C. Dong, "Computer Vision – ECCV 2016" Springer 391-407, 2016

29 C. Dong, "Computer Vision – ECCV 2014" Springer 184-199, 2014

30 L. Zhang, "Adaptive importance learning for improving lightweight image super-resolution network" 128 (128): 479-499, 2020

31 D. P. Kingma, "Adam: a method for stochastic optimization"

32 D. Martin, "A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics" 416-423, 2001