灰度图像的二维Otsu曲线阈值分割法

Otsu法是一个常用的阈值分割方法.为了利用图像的区域信息,本文在二维Otsu法的基础上提出了曲线阈值型Otsu法,传统的二维Otsu法可以看成是该方法的一个特例.实验结果表明,对于含噪图像,它能够获得优于传统二维Otsu法的分割效果.为了减少计算量,提高分割速度,给出了一种递归算法和一种小波变换与递归算法相结合的快速算法.该递推算法只需遍历二维直方图的主对角线和一条次主对角线,与传统Otsu法的递推算法相比,搜索空间由L×L个点减少到2L-1个点.

Two-Dimensional Otsu's Curve Thresholding Segmentation Method for Gray-Level Images

Otsu’s method is a frequently used thresholding segmentation method.In this paper,in order to utilize the information of edge region,we present a curve thresholding segmentation method based on two-dimensional Otsu’s method.The traditional two-dimensional Otsu’s method can be regarded as a special case of it.Experimental results show that for images with noise the proposed method can obtain better segmentation effect than the traditional two-dimensional Otsu’s method.In order to decrease the amount of computation and promote the speed of segmentation,a recursive algorithm and a wavelet transfer combining with the recursive algorithm are shown.The recursive algorithm only need search the main-diagonal and one sub-diagonal of the two dimensional histogram.Compared with the recursive algorithm of traditional two-dimensional Otsu’s method,the searching space is obviously decreased from searching L×L points to 2L-1 points.