一种用于复杂二维图像矩的快速算法  

A NEW ALGORITHMSFOR THE COMPUTATION TWO-DIMENSION IMAGE MOMENTS OF ANY COMPICATED OBJECTS

作　　者：王冰[1] 职秦川[1] 张仲选[1] 耿国华[1] 周明全[1] 

机构地区：[1]西北大学计算机科学系,西安710069

出　　处：《计算机应用与软件》2004年第7期89-92,共4页Computer Applications and Software

摘　　要：本文给出一种新的求矩快速算法。该算法用扫描方法求图像各行各线段的左外边界和右内边界 ;将所有线段转换为标准线段 ,从而使所有可能的线段数目由N2 减少为N ;定义了一组N维数组 ,将求矩过程中大量重复计算的一些算式的结果储存于数组 ,需要时查数组即得 ,从而极大地减少了计算量。该算法原理简单 ,计算结果准确。不同于有些文献给出的算法只适用于无凹图像或不适用于图像中有空洞的情形。本算法适用于任何复杂的有任意多个空洞的图像。从对各种算法求矩运算量比较来看 ,本算法要优于其他算法。In this paper,we present a new method for fast computation geometric moments.First, we consider that the computation of shape moments is summation of the contribution of each row,and the row moments is summation of the contribution of each line-segment.Second,we define a standard line-segment,and each line-segment is transfered to two standard line-segment.So computation line-segment moments is becoming computation standard line-segment.Comparing the number of the line-segment for a N×N image,which is N 2,the number of standard line-segment is N .Third,a group of array is assigned to store the constant,which is repeatedly used during the computation moments.Whenever calculating the moments of a standard line-segment.It can obtain from the array directly.The different from some methods,which are only suitable for convex object or not suitable for the image with holes,our methed is suitable for any complicated shape with any holes.The result of our method is also accurate.Comparison of the computation complexity with some known method is also given,which shows that our algorithm significantly reduces the complextiy.

关 键 词：图像分类 模式识别 计算机视觉 二维图像矩 算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]