A Really Useful Vectorization Algorithm

  • Authors:
  • Dave Elliman

  • Affiliations:
  • -

  • Venue:
  • GREC '99 Selected Papers from the Third International Workshop on Graphics Recognition, Recent Advances
  • Year:
  • 1999

Quantified Score

Hi-index 0.00

Visualization

Abstract

A novel algorithm for the vectorization of binary images is described. It is based on a data structure formed by crack following the outlines of the dark region of the image and applying a heuristic to decide which edges lie along the direction of the vector at any position. The regions are divided into classes, described as strokes and junctions respectively. The idea of a junction as a region containing more than one stroke is introduced, and this is used to inform the process of vectorization in these areas. The quality of the vectors produced compares favourably with those produced by other algorithms known to the author, and the implementation is reasonably efficient, a typical A4 drawing is processed in under ten seconds on my 300 MHz Pentium lap-top.