Novel Algorithm for Extraction of an Object with an Approximately Convex Boundary
Igor Zingman · 2010
One of the fundamental problems in computer vision is the problem of grouping edges to continuous object boundary. We present a robust parameter-free algorithm for extracting the convex boundaries of objects from noisy edges. The algorithm can follow any edge detection scheme which produces a binary edge map and edge directions. The paper describes the algorithm and gives examples of its application to real world images.