eduzhai > Applied Sciences > Engineering >

Rethinking Image Deraining via Rain Streaks and Vapors

  • king
  • (0) Download
  • 20210507
  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: Single image deraining regards an input image as a fusion of a backgroundimage, a transmission map, rain streaks, and atmosphere light. While advancedmodels are proposed for image restoration (i.e., background image generation),they regard rain streaks with the same properties as background rather thantransmission medium. As vapors (i.e., rain streaks accumulation or fog-likerain) are conveyed in the transmission map to model the veiling effect, thefusion of rain streaks and vapors do not naturally reflect the rain imageformation. In this work, we reformulate rain streaks as transmission mediumtogether with vapors to model rain imaging. We propose an encoder-decoder CNNnamed as SNet to learn the transmission map of rain streaks. As rain streaksappear with various shapes and directions, we use ShuffleNet units within SNetto capture their anisotropic representations. As vapors are brought by rainstreaks, we propose a VNet containing spatial pyramid pooling (SSP) to predictthe transmission map of vapors in multi-scales based on that of rain streaks.Meanwhile, we use an encoder CNN named ANet to estimate atmosphere light. TheSNet, VNet, and ANet are jointly trained to predict transmission maps andatmosphere light for rain image restoration. Extensive experiments on thebenchmark datasets demonstrate the effectiveness of the proposed visual modelto predict rain streaks and vapors. The proposed deraining method performsfavorably against state-of-the-art deraining approaches.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×