噪声图像的快速二维Otsu阈值分割

为了提高二维阈值分割法处理速度, 提出了二维Otsu法的快速实现方法。基于二维随机变量的边缘概率分布, 将二维最佳阈值（s*, t*）的求解拆分成两个一维最佳阈值s*和t*的求解; 同时为了改善原算法的分割效果, 引入类内方差的定义, 提出了新的最佳阈值判别式。实验结果表明, 本方法不仅保留了原二维阈值法抗噪性强的特点, 其时间复杂度由O(L⁴)降为O(L), 空间复杂度由S(L²)降为S(L), 且分割错误率低于原二维Otsu法。该方法适合处理高斯噪声图像的快速阈值分割问题。

Fast thresholding algorithm of 2D Otsu for low SNR image

In order to improve 2D thresholding algorithm's processing speed, this paper presented a fast implementation method of 2D Otsu. Based on marginal probability distribution of bivariate discrete random variable, it calculated two 1D optimal threshold, s*and t*, and then kept (s*, t*) as the optimal 2D Otsu threshold. Furthermore, in order to improve the segmentation of the original algorithm, this paper introduced the definition of intra-class variance and proposed a new discriminant. The experimental results show that the proposed algorithm outperforms original algorithm. Without losing the robustness to noise, its time complexity is reduced from O(L4) to O(L), space complexity is reduced from S(L²) to S(L), and the misclassification rate is lower. The method is suitable for handling the fast threshold segmentation of Gaussian noise images.