粒子群优化的多阈值图像自分割算法

Particle Swarm optimization for multilevel thresholding Image self-segmentation method

针对传统的多阈值图像分割方法存在精确度和时间复杂度的问题,提出了粒子群优化的多阈值自分割算法。该方法首先利用自适应滤波函数对图像进行预处理,不仅解决了一维Otsu算法的抗噪性弱的问题,又同时确定了多阈值图像的局部区域类数,然后把改进的粒子群优化算法应用到多阈值图像分割中,对最佳阈值进行搜索优化,改进了图像分割的整体性能。实验结果证明了本文算法的收敛速度与优化质量均优于其他算法,并且这个优越性会随着局部区域类数的增加,体现得更为明显。

Aiming at the problems that the traditional multilevel thresholding Otsu method is inaccurate and costs a lot of time,ascheme for image segmentation multilevel thresholding method based on particle swarm optimization （PSO） is presented. Firstly,thismethod used adaptive filtering function in image preprocessing which could not only solve the weak-noisy defect of one-dimensionalOtsu method,but also define the number of local areas. Then,the improved particle swarm optimization is used to greatly improvethe optimization of the quantity of searching the best thresholds and the efficiency of image segmentation. Results of simulation indi-cated that this algorithm is better than several other algorithms both in convergence rate and quality of optimization. The advantagesare more obvious with the number of the local areas.