【摘要】 提出了一种基于小波惟一描述子的多边形逼近方法。对提取的多边形采用小波描述子 ,通过比较原始多边形和逼近多边形之间小波惟一描述子偏差的大小 ,选择一个最佳的逼近结果 ,以得到一个顶点数递减的近似多边形序列。与现有方法比较 ,本文方法既考虑了轮廓的整体信息 ,又考虑了轮廓的局部信息 ,因而具有更好的鲁棒性。将该方法与面积法及勒让德矩方法进行了比较 ,实验结果表明 ,采用该方法得到的逼近结果具有更好的效果及可靠性更多还原

【Abstract】 A new polygonal approximation method based on the uniqueness wavelet descriptor (UWD) is studied. Firstly, the 1-D periodic signal function of the polygon is obtained, then the differences are calculated between the UWD of the initial contour and those of the approximated polygons with one less vertice, the one with minimum difference is chosen as the appro-ximated polygon, thus a sequence of polygons with different number of vertices is output. By applying the procedure to some digital curves promising results are obtained. Compared with the known methods, such as area based method and Legendre moment based method, experimental results demonstrate that the UWD based algorithm gives better results.更多还原