基于Otsu准则和性质的降维阈值分割法

摘要

by analyzing the traditional two-dimensional Otsu algorithm and two-dimensional linear threshold segmentation method ,It is showed that the two algorithm in exhaustive calculation the calculation amount is very large, the computing time is long, a serious drag slow processing speed, and time requirements of the scene is not applicable.In view of the above situation ,reducing dimensions thresholding segmentation algorithm based on Otsu criterion and property is proposed. This algorithm divides the traditional two-dimensional histogram into two one-dimensional histograms,and calculates the threshold value of image gray value and the threshold value based on neighborhood average gray value by using Otsu criterion and property , then the image segmentation is completed. Through comparison and analysis of the proposed algorithm and the traditional two dimensional Otsu method , the results show that the proposed algorithm can effectively avoid the shortcomings of conventional algorithm in anti noise, while the calculation speed to be faster than that of intercept thresholding method based on two-dimensional Otsu method. As a result , the proposed algorithm is a robust and fast algorithm for threshold segmentation, which is more suitable for real-time applications.%通过对传统二维Otsu算法和二维直线阈值分割法进行分析,发现该两种算法在穷举计算时其计算量很大,计算时间长,严重拖慢了处理速度,不适用与时间要求高的场景.所以本文提出了一种基于Otsu准则和性质的降维快速阈值分割法.该算法将传统的一维直方图转换成两个一维直方图,再根据Otsu准则和性质快速计算出其基于图像灰度值和基于邻域平均灰度值的两个阈值,并使用以上两个阈值和二维直方图数据完成图像二值化.通过对本文的算法与二维Otsu直方图法以及二维阈值直线法进行数据比较,其结果表明:本文的算法可以有效避免传统算法在噪声干扰方面的缺陷,同时运算速度大大被提高,更加适用于实时图像处理.