基于差分法的灰度图像矩的快速算法  

A NEW COMPUTATION OF GEOMETRIC MOMENTS FOR GRAY LEVEL IMAGE BASED ON DIFFERENTIAL IMAGE

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

机构地区：[1]西北大学计算机科学系,陕西西安710069

出　　处：《计算机应用与软件》2005年第11期17-19,111,共4页Computer Applications and Software

基　　金：国家自然科学基金资助项目(60271032)。

摘　　要：本文提出了一种新的对于灰度图像的几何矩的快速算法。首先运用图像差分法,将图像函数f(x,y)变换为图像函数d(x,y)。其次,从xn(n=1,2,3)的递推求和得到一组数组。灰度图像的几何矩可以由该数组和函数d(x,y)计算获得。这种方法的优点在于:图像行(列)中具有相同像素值的连续部分,经差分后,除端点外的其它部分都为0,求矩无需考虑值为0的像素。所以,求矩计算量大大地降低了。文中给出了实验结果,和其它灰度图像求矩算法相比,文中算法在大多数情形下都极大地降低了计算复杂度。该算法乘法和加法的运算次数大约是Belkasim’s算法的47.4%和59.8%,大约是Yang’s算法的35%和51.8%。The paper proposes a novel approach to calculate geometric moment for gray level image. Firstly the technique of differential image is involved to transform the image fromf（x,y） to d（x,y）. Secondly a group of array is generated from the iterative sum of x^n（ n = 1,2, 3）. Geometric moments can be computed from d（x,y） and the array. The advantage of the method is that the number,which d（x,y） is zero, may increase greatly. It is not require to deal with the d（x,y） which is zero, so computation cost is decreased greatly. Some tests are given using the method. Compared with some known method to calculate geomertric moments for gray level image, which shows that our algorithm can reduce computational complexity significantly in mose cases. The multiplications and additions is about 47.4% and 59.8% with Belkasim＇ s algorithm,is about 35% and 51.8% with Yang＇s algorithm respectively.

关 键 词：差分法 灰度图像矩 快速算法 图像函数 计算原理 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]