基于灰色系统理论的阈值自动选取算法

Automatic threshold selection based on grey system theory

针对阈值方法常需要人工干预的问题,提出了一种基于灰色系统理论的阈值自动选取算法。首先利用降低灰度级后的直方图检测峰值,然后自动采集峰间内侧附近的样本作为灰色预测的种子点。通过灰色理论GM(1,1)模型预测种子点发展走向,并计算模拟交汇点,得到最优阈值。利用该算法与经典的Otsu,Kapur算法以及文献[3]和[4]中的方法对15组不同复杂度图像进行对比阈值分割,并采用AER进行分割评估,实验表明新算法平均分割误差为19.37%,低于上述四种方法。

It is a tricky problem for threshold techniques to decide an optical threshold automatically.An automatic threshold selection method based on gray system theory is proposed in this paper to calculate the optical threshold without manual intervention.By analyzing the degraded gray-level histogram,the samples（seed points） around medial of the peaks are obtained.Then the novel method applies GM（1,1） model to predict the trend of the seed points.Finally,the intersection is calculated, which acts as the optimal threshold.15 images are segmented by the proposed method compared with traditional methods（i.e.Otsu,Kapur） and two improved methods in references[3] and [4].The segmentation results are evaluated by area error rate.The average error rate is 19.37% on a data set with diverse segment complexity.The results show that the proposed method can segment objects from images effectively.