eduzhai > Applied Sciences > Engineering >

Multi-Domain Image Completion for Random Missing Input Data

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 31 pages

Abstract: Multi-domain data are widely leveraged in vision applications takingadvantage of complementary information from different modalities, e.g., braintumor segmentation from multi-parametric magnetic resonance imaging (MRI).However, due to possible data corruption and different imaging protocols, theavailability of images for each domain could vary amongst multiple data sourcesin practice, which makes it challenging to build a universal model with avaried set of input data. To tackle this problem, we propose a general approachto complete the random missing domain(s) data in real applications.Specifically, we develop a novel multi-domain image completion method thatutilizes a generative adversarial network (GAN) with a representationaldisentanglement scheme to extract shared skeleton encoding and separate fleshencoding across multiple domains. We further illustrate that the learnedrepresentation in multi-domain image completion could be leveraged forhigh-level tasks, e.g., segmentation, by introducing a unified frameworkconsisting of image completion and segmentation with a shared content encoder.The experiments demonstrate consistent performance improvement on threedatasets for brain tumor segmentation, prostate segmentation, and facialexpression image completion respectively.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×