Laser & Optoelectronics Progress, Volume. 58, Issue 4, 0410024(2021)

Image Deraining Algorithm via Multiflow Expansion Residual Dense Network

Weiwei Wang1, Yayu Zhai2, Ping Chen1、*, and Fengcai Cao1
Author Affiliations
  • 1School of Information and Communication Engineering, North University of China, Taiyuan, Shanxi 0 30051, China
  • 2China Institute of Marine Technology&Economy, Beijing 100081, China;
  • show less
    Figures & Tables(11)
    Overall framework of image rain removal algorithm
    Multiflow expansion residual dense network
    Examples of image decomposition. (a) Rainy image; (b) base layer of rainy image; (c) base layer of rainless image; (d) detail layer of rainy image
    Color curves of base layer. (a) Rainless image; (b) rainy image
    Example of residual image. (a) Rainless image; (b) rainy image; (c) residual image
    Visual comparison of different algorithms on composite rainy images. (a1)(a2) Rainy image; (b1)(b2) GMM algorithm; (c1)(c2) DSC algorithm; (d1)(d2) DCN algorithm; (e1)(e2) DID-MDN algorithm; (f1)(f2) MERDN algorithm
    Visual comparison of different algorithms on real rainy images. (a1)(a2) Rainy image; (b1)(b2) GMM algorithm; (c1)(c2) DSC algorithm; (d1)(d2) DCN algorithm; (e1)(e2) DID-MDN algorithm; (f1)(f2) MERDN algorithm
    • Table 1. Detailed setting of parameter layer

      View table

      Table 1. Detailed setting of parameter layer

      LayerFilter sizeDilated factorNumber of filters
      Convolution×13×3132
      Convolution×13×3232
      Convolution×13×3332
      Dense block×91×1164
      Dense block×93×3232
      Convolution×13×313
    • Table 2. PSNR of different algorithms unit: dB

      View table

      Table 2. PSNR of different algorithms unit: dB

      AlgorithmDataset 1Dataset 2Dataset 3
      GMM23.7528.3630.70
      DSC22.0324.1630.02
      DCN21.9728.1729.42
      ID-CGAN23.8123.3920.78
      DID-MDN26.1728.2728.90
      MERDN26.4628.8131.90
    • Table 3. SSIM of different algorithms

      View table

      Table 3. SSIM of different algorithms

      AlgorithmDataset 1Dataset 2Dataset 3
      GMM0.78440.87120.8928
      DSC0.79850.86790.8679
      DCN0.83330.91340.9033
      ID-CGAN0.80720.82750.8519
      DID-MDN0.84090.86250.8762
      MERDN0.88410.91730.9381
    • Table 4. Efficiency comparison of different algorithms

      View table

      Table 4. Efficiency comparison of different algorithms

      AlgorithmGMMDCNDID-MDNMERDN
      Time /s89729115
    Tools

    Get Citation

    Copy Citation Text

    Weiwei Wang, Yayu Zhai, Ping Chen, Fengcai Cao. Image Deraining Algorithm via Multiflow Expansion Residual Dense Network[J]. Laser & Optoelectronics Progress, 2021, 58(4): 0410024

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Image Processing

    Received: Sep. 25, 2020

    Accepted: Oct. 14, 2020

    Published Online: Feb. 22, 2021

    The Author Email: Chen Ping (pc0912@163.com)

    DOI:10.3788/LOP202158.0410024

    Topics