基于双界线的Otsu阈值分割法及其快速算法

传统2维Otsu阈值分割法由于运算时间长、抗噪能力不足而在应用中受到限制。为了克服这些缺点，提出了一种基于双界线的2维Otsu阈值理论及其快速算法。在新的2维直方图中，两条平行于对角线的界线决定目标和背景区域的宽度，垂直于对角线的分割直线决定阈值大小。该算法运用Roberts算子和线性拟合法确定双界线，然后运用改进的Otsu法计算最佳阈值，最后对噪声区域进行后处理。实验结果表明，该算法不仅运算速度快，而且具备较好的分割质量和抗噪性能。同时，快速算法的引入，进一步降低了运算量，使得该算法具备更好的实时性。

Otsu thresholding segmentation method based on two boundaries and its fast algorithm

The traditional two-dimensional(2D) Otsu thresholding segmentation algorithm has been restricted in application for long-paying computation and poor anti-noise capacity. In order to overcome these drawbacks, this paper proposed a 2D Otsu method based on two boundaries and its fast algorithm. In new 2D histogram, two boundaries parallel to diagonal decided the width of object and background region, and the segmentation line perpendicular to diagonal decided the value of threshold. The proposed method applied Roberts operator and Linear fitting to obtain two boundaries, then utilized modified Otsu algorithm to calculate threshold, finally reprocessed noise areas. Experimental results show that, proposed method not only has a fast computational speed but also owns better segmentation quality and anti-noise capacity. Meanwhile, the proposed fast algorithm lowers the computational complexity further, which makes the real-time of proposed method better.