二维Otsu阈值法的快速算法

Fast algorithm for 2D Otsu thresholding algorithm

最大类间方差法(Otsu)是图像分割的经典算法,在其基础之上发展起来的二维Otsu阈值分割法由于计算复杂而制约了其应用。针对这一缺点,提出一种改进的二维Otsu阈值法的快速算法。首先将原始二维直方图划分成M×M个区域,将每个区域视为1个点,构造新的二维直方图,在其上利用二维Otsu以及快速递推算法,得到分割阈值所处的区域编号;既而对所确定的区域再次使用二维Otsu算法得到原始图像的分割阈值。实验结果证明,改进算法有效地提高了计算速度,降低了算法的空间复杂度,且分割效果与原始算法基本一致。

Otsu algorithm is widely used in classic image segmentation, while the application of the two-dimensional Otsu thresholding algorithm based on the Otsu algorithm has been restricted for the complex computation. Concerning this, this paper proposed an improved two-dimensional Otsu thresholding algorithm. The authors first divided the two-dimensional histogram into regions, and took each region as a point to form a new two-dimensional histogram, to which 2D Otsu thresholding algorithm and the fast reeursive algorithm were applied, getting the region number of the threshold. Then the two algorithms were applied again on the region and finally the threshold for the original image was obtained. The experimental results show that the proposed algorithm greatly reduces the running time and the storage space, and gets basically the same results as the original algorithm.