Contour smoothing algorithm based on contour extremes

Abstract

In this article we propose linear time algorithm for contour smoothing, based on finding extremes. First we find vertexes where contour convexity changes, than obtain local minimums, maximums and points of support, which should be used in resulting contour. The main goal of proposed approach is to compute accurate interior area of a bounding contour, it was successfully applied for recovering object contour after segmentation algorithms or human annotations, for contours noise reduction after jpeg compression.