基于改进粒子群的二维模糊散度多阈值图像分割

TWO-DIMENSIONAL FUZZY DIVERGENCE MULTI-THRESHOLD IMAGE SEGMENTATION BASED ON IMPROVED PSO

针对只利用灰度信息的图像分割算法精度低和未加优化的二维多阈值分割算法耗时长,以及粒子群算法易出现虚假收敛的问题,提出基于改进粒子群的二维模糊散度多阈值图像分割算法。考虑图像像素空间邻域信息,建立二维隶属度函数,进而推导出二维多阈值α型模糊散度作为选取最佳阈值的准则函数;用线性递减和线性递增函数分别对粒子群算法的自我认知和社会认知部分做改进;用改进粒子群算法优化求解二维多阈值α-型模糊散度的多组阈值。实验结果表明,该算法可以提高分割精度和改善分割性能。

Aiming at the low accuracy of image segmentation algorithm only using gray level information and the time-consuming of two-dimensional multi-threshold segmentation algorithm without optimization,and considering the problem that the PSO is prone to false convergence,we propose a two-dimensional fuzzy divergence multi-threshold image segmentation algorithm based on improved PSO.Considering the image pixel spatial neighborhood information,the two-dimensional membership function was established,and then the two-dimensional multi-thresholdα-type fuzzy divergence was derived as the criterion function for selecting the optimal threshold;the self-cognition and social cognition parts of the PSO were improved by linear decreasing function and linear increasing function respectively;we used the improved PSO to optimize the multi-group thresholds of two-dimensional multi-thresholdα-type fuzzy divergence.The experimental results show that our algorithm can improve the segmentation accuracy and the segmentation performance.