图像目标外接多边形及凸壳的一种构造方法

A Method of Constructing Bounding Polygons and Convex Hulls of Image Objects

对二值图像进行Hough变换后,在(ρ,θ)空间中选取了一组边界对应点,通过计算与这些边界对应点对应的图像空间中直线的交点,构造了图像目标的外接多边形;通过比较相距π/2rad的投影区间长度是否相等,或区间长度的乘积是否为最小,得到了形状外接正方形和外接最小面积矩形;利用构造形状外接多边形的方法并通过增加边的数目,构造了形状的近似凸壳.实验和理论分析表明,文中算法具有好的抗噪性能和广泛的适用范围.

After a binary image is transformed by the Hough transform under the （p, θ） line parameterization, a set of pairs of boundary points in the （p,θ） space is chosen. By computing the intersecting points of the lines in the image space, which are determined by the chosen pairs, the bounding polygon of the image objects is obtained. In addition, by determining those pairs of boundary point pairs in the （p, θ） space, where the θ difference of two boundary point pairs is π/2 rad but the p lengths of the two pairs are equal, or their product is minimum, the bounding square or the minimum bounding rectangle are respectively obtained in the same way as for the bounding polygon construction. Furthermore by increasing the number of the edges in the above bounding polygon construction process, an approximate convex hull of object is achieved. Both experimental results and analysis show that our proposed algorithms have good noise-resistance ability and a large range of applicability.