快速最大熵多阈值图像分割算法

A FAST SEGMENTATION METHOD FOR MULTI-THRESHOLD IMAGE BASED ON MAXIMUM ENTROPY

阈值方法是一种重要的图像分割方法,在图像分割中得到了广泛的应用。最大熵算法虽然是图像分割阈值法中较好的方法之一,但是,由于传统的最大熵算法通常用穷举法求解,使得处理多阈值问题时运算速度太慢,难以满足应用需求。为了快速有效地确定阈值,提出一种改进的最大熵算法。通过递推公式将穷举法求解过程中需要重复计算的变量,预先计算后存入二维表备用,使整体计算量减少了一个数量级。通过对测试图像的分割实验,表明该算法与传统的最大熵算法相比运算速度有非常显著的提高,能够满足一般的应用需求。

Thresholding is an important and wide-applied image segmentation method.Though maximum entropy algorithm is one of the good methods in thresholding of image segmentation,however,the processing of multi-thresholds issue suffers from extremely low computation speed due to the brute-force method usually used by conventional maximum entropy algorithm in solving problems,it is difficult to meet the application requirements.In order to fast and effectively determine the thresholds,in this paper,a modified maximum entropy method is proposed.It enables the whole computation complexity to have a reduction of an order of magnitude.This is accomplished by a recursive formula which pre-calculates some variables to be repeatedly computed during solving process of brute-force method and stored them in a 2D look-up table as the standby.The segmentation experiment on testing images show that the computational speed of the proposed method has significantly improved compared with conventional maximum entropy method and is able to accommodate general application requirements.