基于回溯搜索优化算法的图像分割

Image Segmentation Based on Backtracking Search Optimization

阈值法是一种简单且有效的图像分割技术。但是随着阈值数目的增加,求解阈值的计算量增大并且实时性降低,这给多阈值图像分割带来了很大的困难。为了克服这一困难,把多阈值分割看作一个优化问题。将最大类间方差法当作要优化的函数,用回溯搜索优化算法求解要优化的函数,从而实现多阈值图像分割。将提出的多阈值算法应用于基准测试图像上,并与传统阈值分割法比较。实验结果表明:回溯搜索优化算法很好的解决了最大类间方差算法求解多阈值分割实时性差的问题,并验证了该算法应用在图像分割上是可行的。

Threshold method is a simple and effective image segmentation technology.However,with the increase of the number of thresholds,the computational complexity of solving thresholds increases and the real-time performance decreases,which brings great difficulties to multi-threshold image segmentation.In order to overcome this difficulty,multi-threshold segmentation is considered as an optimization problem.The maximum inter-class variance method is regarded as the function to be optimized,and the backtracking search optimization algorithm is used to solve the function to be optimized,so as to achieve multi-threshold image segmentation.The proposed multi-threshold algorithm is applied to the benchmark image and compared with the traditional threshold segmentation method.The experimental results show that the retrospective search optimization algorithm solves the problem of poor real-time performance of the maximum interspecific variance algorithm in multi-threshold segmentation,and verifies that the algorithm is feasible in image segmentation.