• DocumentCode
    3398347
  • Title

    Skeletonization using the divergence of an anisotropic vector field flow

  • Author

    Giuliani, Donatella

  • Author_Institution
    Univ. of Bologna, Bologna, Italy
  • fYear
    2013
  • fDate
    23-25 Oct. 2013
  • Firstpage
    1
  • Lastpage
    9
  • Abstract
    In this paper we introduce a procedure to obtain both the automatic positioning of an initial contour for edge extraction and a new approach to skeletonize the captured contour. We propose a parametric deformable method that relies on a generalized anisotropic flow for the evaluation of an external force field which will be applied on active contour procedure. Therefore we recur to field divergence to analyze its convergence, in order to better place an initial curve. We put in evidence that the divergence of the vector field satisfies an anisotropic diffusion equation as well. The curves of positive divergence may be thought as propagating fronts of an evolution function. It has been proved that the sets of points in the image domain where divergence assumes positive values, converge to the skeleton of the extracted contour. We suggest that the divergence of a vector field may be an alternative medial axis function as soon the steady configuration of the anisotropic flow has been reached.
  • Keywords
    computational geometry; convergence; edge detection; feature extraction; image thinning; active contour procedure; anisotropic diffusion equation; anisotropic vector field flow; automatic contour positioning; captured contour skeletonization; contour extraction; convergence analysis; edge extraction; external force field evaluation; generalized anisotropic flow; image domain; medial axis function; parametric deformable method; positive divergence curves; positive values; vector field divergence; Equations; Force; Image edge detection; Mathematical model; Shape; Skeleton; Vectors; GGVF; MAT; anisotropic flow; divergence; medial axis; skeleton;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Applied Imagery Pattern Recognition Workshop (AIPR): Sensing for Control and Augmentation, 2013 IEEE
  • Conference_Location
    Washington, DC
  • Type

    conf

  • DOI
    10.1109/AIPR.2013.6749321
  • Filename
    6749321