eduzhai > Applied Sciences > Engineering >

Retinal Image Segmentation with a Structure-Texture Demixing Network

  • king
  • (0) Download
  • 20210506
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Retinal image segmentation plays an important role in automatic diseasediagnosis. This task is very challenging because the complex structure andtexture information are mixed in a retinal image, and distinguishing theinformation is difficult. Existing methods handle texture and structurejointly, which may lead biased models toward recognizing textures and thusresults in inferior segmentation performance. To address it, we propose asegmentation strategy that seeks to separate structure and texture componentsand significantly improve the performance. To this end, we design astructure-texture demixing network (STD-Net) that can process structures andtextures differently and better. Extensive experiments on two retinal imagesegmentation tasks (i.e., blood vessel segmentation, optic disc and cupsegmentation) demonstrate the effectiveness of the proposed method.

Please select stars to rate!


0 comments Sign in to leave a comment.

    Data loading, please wait...