基于迭代与优化搜索策略的二维Otsu算法

Two-Dimensional Otsu Algorithm Based on Iteration and Optimizing Search Strategy

针对二维Otsu阈值算法计算复杂度高的缺点,提出一种快速算法.该方法首先对阈值的判定公式作了一定的推导以减少计算变量,然后用迭代的方法消除计算冗余,最后结合小波分解和对角线优先搜索的策略获得图像分割阈值.结果表明,快速算法比传统二维Otsu法的计算量减少了4个数量级,相对于基于迭代的查询表算法,其运行时间也减少了50%以上,而各种算法得到的图像分割效果差异不大.

This paper proposes a fast algorithm for two-dimensional Otsu threshold algorithm that overcomes the disadvantage of computational complexity. Firstly,in order to reduce the variables for calculation,a formula to determine threshold is derived. Secondly,an iterative method is used to eliminate redundant computation. Finally,using wavelet decomposition in combination with the diagonal firstsearch strategy,the threshold of image segmentation is obtained. Compared with original two-dimensional Otsu algorithm,the computation of the proposed fast algorithm is reduced by 4 orders of magnitude and,compared with iteration-based query algorithm,its computational time is reduced by 50% or more. The results of image segmentation based on the above-mentioned algorithms are barely different.