基于微粒群算法的灰度图像阈值分割的改进

Improvement of grayscale image threshold segmentation based on PSO algorithm

为了充分利用灰度图像的灰度信息和空间信息,提高分割精确度和最优阈值的求解速度,提出一种基于微粒群算法的阈值分割方法——PSO-SDAIVE算法。该算法对传统的二维直方图进行改进,生成差值属性灰度直方图,同时对灰度均值和二维熵的计算进行改进,生成空间差值属性信息值熵(SDAIVE),最后用微粒群算法搜索SDAIVE的最大值。在实验中,对头部CT图像进行分割,实验结果表明,这种分割方法能精确地获得分割阈值,并有很好的抗噪声能力,节省计算时间。

To fully utilize gray information and spatial information of grayscale,increase the segmentation precision and the solution speed of optimal threshold,a threshold segmentation method based on PSO algorithm is proposed.This algorithm is called PSO-SDAIVE algorithm.At first,the traditional 2D histogram is improved and the 2D D-value attribute gray histogram is created.Otherwise,the computation of gray average and 2D entropy are improved.The improved entropy is called spatial D-value attribute information value entropy（SDAIVE）.At last,simple PSO algorithm is used to solve the SDAIVE maximum.In experiment,head CT images are segmented.Experimental results show that this new method can accurately obtain segmentation threshold.Otherwise,this method has strong anti-noise capability and saves computation time.