eduzhai > Applied Sciences > Engineering >

Video compression with low complexity CNN-based spatial resolution adaptation

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 7 pages

Abstract: It has recently been demonstrated that spatial resolution adaptation can beintegrated within video compression to improve overall coding performance byspatially down-sampling before encoding and super-resolving at the decoder.Significant improvements have been reported when convolutional neural networks(CNNs) were used to perform the resolution up-sampling. However, this approachsuffers from high complexity at the decoder due to the employment of CNN-basedsuper-resolution. In this paper, a novel framework is proposed which supportsthe flexible allocation of complexity between the encoder and decoder. Thisapproach employs a CNN model for video down-sampling at the encoder and uses aLanczos3 filter to reconstruct full resolution at the decoder. The proposedmethod was integrated into the HEVC HM 16.20 software and evaluated on JVET UHDtest sequences using the All Intra configuration. The experimental resultsdemonstrate the potential of the proposed approach, with significant bitratesavings (more than 10 ) over the original HEVC HM, coupled with reducedcomputational complexity at both encoder (29 ) and decoder (10 ).

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×