正余弦优化算法在多阈值图像分割中的应用

Application of Sine Cosine Optimization Algorithm to Multi-threshold Image Segmentation

阈值的选取对图像分割后的效果有重要的影响,在传统的图像分割中存在分割结果单一,灵活度不强,以及容易陷入局部最优的问题。为了确定图像分割的最佳阈值,本文针对图像分割过程中涉及的阈值选取问题,提出一种基于正余弦优化算法(sine cosine algorithm,SCA)的多阈值图像分割方法。该算法以最大类间方差作为正余弦算法的适应度函数,通过正弦函数和余弦函数的变化来更新当前解在每一维度上的位置,候选解利用多个随机算子围绕最优解进行正余弦的波动来完成每一次的寻优过程,通过迭代计算更新最优解的位置,从而确定图像分割的最佳阈值。选取4幅标准测试图像进行实验,通过与粒子群优化算法进行峰值信噪比、结构相似法和寻优时间3方面的对比,结果表明:将正余弦优化算法应用在图像分割中可以获得更准确的分割阈值和更高的分割效率,具有很强的实用性。

Threshold selection has an important impact on the effect of image segmentation.In traditional image segmentation,there are some problems such as single segmentation result,weak flexibility and easy to fall into local optimum.In order to determine the optimal thresholds for image segmentation,a multilevel-thresholding image segmentation method based on sine cosine algorithm (SCA) is proposed in this paper.The method takes the between-class variance as the fitness function of the sine-cosine algorithm,updates the position of the current solution in each dimension by changing the sine function and cosine function.The candidate solution uses multiple random operators to carry out sine-cosine fluctuation around the optimal solution to complete each optimization process,and updates the optimum through iterative calculation.The optimal thresholds of image segmentation are determined by the location of the solution.Four standard test images are selected for experiment and compared with PSO algorithm in peak signal to noise ratio,structural similarity method,and optimization time.It shows the application of sine-cosine algorithm in image segmentation can obtain more accurate threshold and higher segmentation efficiency,which has a strong practicability.