A Non-local Measure for Mesh Saliency via Feature Space Reduction

Gençtav A., Gençtav M., Tari S.

in: Research in Data Science, Ellen Gasparovic,Carlotta Domeniconi, Editor, Springer, Cham, Zug, pp.167-175, 2019

  • Publication Type: Book Chapter / Chapter Research Book
  • Publication Date: 2019
  • Publisher: Springer, Cham
  • City: Zug
  • Page Numbers: pp.167-175
  • Editors: Ellen Gasparovic,Carlotta Domeniconi, Editor
  • TED University Affiliated: No


Using data analysis tools, we present a mesh saliency measure in the form of one-parameter family of functions that depends on both local and global factors. The preliminary results seem to agree with intuition. As the parameter increases, consistent with the classical view on the subject, the measure attains its peak values around curvature extrema. Local to global integration is achieved in two steps: the first step is a clever feature space construction, and the second is a dimensionality reduction via proper choice for matrix decomposition. The presented work is an interesting application of using mathematical techniques of data science in the scope of visual perception.