改进的三维Otsu阈值分割法及其快速实现

Improved 3-D Otsu's Method and Its Fast Implementation

常用的三维直方图一般由灰度级-平均灰度级-中值灰度级构成,其区域划分方式存在明显的不足。对此,本文提出了一种改进的三维Otsu法。该方法基于灰度级-平均梯度-中值梯度建立三维直方图,采用新的区域划分方式提高划分精度;为提高算法效率,推导了该方法的快速迭代公式。测试实验表明:与传统三维Otsu法相比,该方法的计算效率高,分割精度好、抗噪性强,表现出较好的应用价值。

The existing 3-D histogram is composed by gray level, average gray level and median gray level. Its region division has obvious shortcomings. So this paper presents an improved 3-D Otsu＇s method. The proposed method uses a new histogram composed by gray level, average gradient and median gradient and its regional division method. In order to improve the computational efficiency, we have deducted its fast iterative algorithm. Experiments results show that compared with the traditional 3-D Otsu＇s method, the proposed method is better in the computational efficiency, the segmentation accuracy anti-noise ability, showing a good application value.