eduzhai > Applied Sciences > Engineering >

Wavelet Channel Attention Module with a Fusion Network for Single Image Deraining

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 5 pages

Abstract: Single image deraining is a crucial problem because rain severely degeneratesthe visibility of images and affects the performance of computer vision taskslike outdoor surveillance systems and intelligent vehicles. In this paper, wepropose the new convolutional neural network (CNN) called the wavelet channelattention module with a fusion network. Wavelet transform and the inversewavelet transform are substituted for down-sampling and up-sampling so featuremaps from the wavelet transform and convolutions contain different frequenciesand scales. Furthermore, feature maps are integrated by channel attention. Ourproposed network learns confidence maps of four sub-band images derived fromthe wavelet transform of the original images. Finally, the clear image can bewell restored via the wavelet reconstruction and fusion of the low-frequencypart and high-frequency parts. Several experimental results on synthetic andreal images present that the proposed algorithm outperforms state-of-the-artmethods.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...