物体轮廓线的多边形拟合

本文讨论了在给定数字轮廓线顶点数目为N 的情况下，从其上选择k 个点来构造拟和
多边形来近似原图，使得物体轮廓线的形状丢失最小。本文的贡献在于1)本文的轮廓线拟和是以物体轮廓线的形状的信息丢失最小为目标的。而传统的方法是以拟和多边形和原图的面积差距最小。2）本文认为每一个点具有一定的形状信息，然后建立了一个以丢失的形状信息最小的0-1 规划模型; 3)使用了两种方法对轮廓线进行了多边形拟合，并把两个结果进行了比较。
关键词： 初始轮廓线; 信息论;0-1 规划模型 动态规划
Abstract: This paper presents the problem of polygonal approximation of digitized outlines in
which k polygonal is constructed by selecting k vertexes from original outlines so that the lost
shape information of outline could be minimized. The contributions of this paper are 1) The
polygonal approximation here targets to minimized the lost shape information rather than the lost area information which is used in traditional way; 2) A 0-1 mathematical model which targets to minimize the lost shape information is build, since each vertex in the outline represents the shape information; 3) Last, two means are used to approximate the digitized outline and the results are compared.
Keywords: outline; 0-1 programming model; dynamic programming