Novel Contour Vectorization Using Holistic Feature of Object

In this paper, a novel contour vectorization approach based on holistic feature of object is proposed, which aims at avoiding segmenting an entire contour of target into some discrete sets of curves. This method consists of two steps:(1) Digitize pixel-based contour into vectors or arcs referring to holistic feature of object, and the contour trend used in the implementation of vectorization is defined here;(2)Approximate vectors and arcs to regular figures, including various postprocessing treatments. Innovative and significant features of this approach includes: preservation of original contour shape; low-loss of performance in noisy environments; linear time complexity and neglectable space complexity. The experimental results obtained confirm the validity of the proposed approach.