eduzhai > Applied Sciences > Engineering >

Blur-Attention A boosting mechanism for non-uniform blurred image restoration

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 32 pages

Abstract: Dynamic scene deblurring is a challenging problem in computer vision. It isdifficult to accurately estimate the spatially varying blur kernel bytraditional methods. Data-driven-based methods usually employ kernel-freeend-to-end mapping schemes, which are apt to overlook the kernel estimation. Toaddress this issue, we propose a blur-attention module to dynamically capturethe spatially varying features of non-uniform blurred images. The moduleconsists of a DenseBlock unit and a spatial attention unit with multi-poolingfeature fusion, which can effectively extract complex spatially varying blurfeatures. We design a multi-level residual connection structure to connectmultiple blur-attention modules to form a blur-attention network. Byintroducing the blur-attention network into a conditional generationadversarial framework, we propose an end-to-end blind motion deblurring method,namely Blur-Attention-GAN (BAG), for a single image. Our method can adaptivelyselect the weights of the extracted features according to the spatially varyingblur features, and dynamically restore the images. Experimental results showthat the deblurring capability of our method achieved outstanding objectiveperformance in terms of PSNR, SSIM, and subjective visual quality. Furthermore,by visualizing the features extracted by the blur-attention module,comprehensive discussions are provided on its effectiveness.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×