Otsu阈值分割法特点及其应用分析

Analysis of characteristics and application of Otsu threshold method

Otsu方法是应用最广泛的阈值分割法之一,对直方图呈双模或者多模分布的图像分割效果好,直方图呈单模分布的分割效果差。针对这个结果,从Otsu分割法理论出发,分析其分割阈值的特点,证明得出:Otsu单阈值分割时最佳阈值为该阈值分割出的两类灰度均值的平均,Otsu多阈值分割时的多个最佳阈值是这些阈值分割出的多类中相邻两类灰度均值的平均。单阈值Otsu法用于表面缺陷检测图像分割时将大部分背景错分为目标,对此,提出了一种改进的Otsu阈值分割。实验结果表明,改进的算法分割效果优于传统算法,在比较准确分割目标的同时有效减少背景的误分割。

Otsu method is one of the most widely used approaches used to decide the threshold for a satisfactory result when the image histogram is bimodal or multimodal distribution,but it fails when the histogram of an image is unimodal.In view of this result,we analyze the characteristics of threshold based on the Otsu segmentation theory.This paper draws a conclusion that the single threshold of the Otsu method is the average of the means of two classes partitioned by this threshold,and multi-threshold of the Otsu method is the average of the means of adjacent two classes portioned by these thresholds.The background is misclassified to object when the single threshold of Otsu method is used for image segmentation in the application of surface defect detection.To address this problem,an improved Otsu method is proposed,and the experiment result shows that the improved algorithm is better than traditional Otsu method.It can not only accurately segment the target from the background but also effectively reduce the false alarm.