二维Otsu阈值分割算法的改进及应用  

Improvement of Two-dimensional Otsu Algorithm and Its Application

作　　者：王永波[1] 陈继荣[1] 

机构地区：[1]中国科学技术大学电子工程与信息科学系,安徽合肥230027

出　　处：《计算机仿真》2008年第4期263-266,共4页Computer Simulation

摘　　要：为改善传统二维Otsu阈值分割算法处理图像的计算复杂度高、实时性差、受噪声干扰严重等缺点,提出一种改进的快速二维Otsu阈值自动分割算法。通过改变二维直方图判别域的划分将二维阈值求解转化为一维阈值求解,对原图像的直方图采用二分法得到最优分割阈值的初始值,然后在此基础上通过两个分量方向上迭代求解得到图像的最优分割阈值。理论分析与仿真车牌实验表明,与传统二维阈值分割算法相比,此算法的计算复杂度远远低于原始二维Otsu算法,且分割效果和原始算法的基本一致。The traditional two - dimensional Otsu threshold algorithm is not a good method for processing the images due to the high computation complexity and noise sensitivity. To solve this problem, an improved two - dimensional Otsu threshold automatic segmentation algorithm is proposed. The algorithm changes the area partition method to transfer the two - dimensional threshold solving problem into one - dimensional threshold＇s, gets the initial value of optimal threshold value by using dichotomy method for the histogram of original image, and then computes the optimal threshold value by iterated method at the two components direction. The theoretical analysis and the results of emulational license plate experiments show that the computational time of the new method is far less than that of the original one and the effect of segmentation is almost the same as using the original one.

关 键 词：图像分割 最大类间方差法 车牌图像 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]