• Cartoon-versus-texture issue
  • Image decomposition into cartoon and texture patterns plays an important role in high definition image processing. In order to improve the definition of an image or a video, consider (a) decomposing the image into smooth/texture components, (b) enhancing every component separately and then (c) fusing the enhanced results. This yields high quality images and videos, even for images issued from a low definition camera.

    How to perform cartoon-texture decomposition?
    Reference [Atto & Mercier, 2015] provides functional objectives that operate structural image decomposition associated with sharp cartooning effect (piecewise smooth fields separated by 2D curves) and rich textural information (sparse variation fields).

    Sparseness/Compressibility
    In addition with the image definition improvment (enhancing separately cartoon and texture features), cartoon-versus-texture decomposition issue is another way of addressing image compression problem: cartoon is highly compressible (sparse in the wavelet domain) whereas texture compressibility depends only on texture richness ('Disney' texture is more sparse than 'Iguana' texture).
  • Iguana
    Iguana image

  • = Cartoon - Iguana <+> Texture - Iguana (skin)
  • 'Disney'
    Iguana image

  • = Cartoon - Iguana <+> Texture - Iguana (skin)