改进的2维Otsu法及混沌粒子群递推的阈值分割

鉴于现常用的灰度级-平均灰度级2维直方图区域划分将部分目标和背景点错分成边缘和噪声点这一不足，为此提出了一种基于灰度级-梯度2维直方图的Otsu阈值选取新方法，利用混沌粒子群优化算法来寻找分割阈值，并提出在迭代过程中，采用递推方法来大大减少适应度函数的重复计算。实验结果表明，与最近提出的基于灰度级-平均灰度级2维直方图Otsu法及粒子群的快速图像分割方法相比，该新方法由于尽可能地考虑了所有目标点和背景点，从而使分割后的图像区域内部均匀、边界形状准确、特征细节清晰，同时运行时间几乎不到现有算法的1/3，而且粒子群处理的收敛精度得到了进一步提高。

Thresholding Based on Improved 2D Otsu Method and Chaotic Particle Swarm Optimization

In view of the shortage of regional division of the commonly used gray level-average gray level two-dimensional histogram, which some object and background inner points are wrongly divided as edge and noise points, an improved Otsu threshold selection method based on gray level-gradient two-dimensional histogram is proposed in this paper. The chaotic particle swarm algorithm is used to search for the best threshold. The repeat computations of the fitness function in iteration are reduced significantly using recursion. Compared with fast image segmentation algorithm based on gray level-average gray level 2-D Otsu method and particle swarm optimization , the experimental results show that the algorithm proposed in this paper not only considers all the object and background inner points and achieves a good segmentation quality in uniform regions, accurate borders and clear details of features,but also the running time is reduced to only 1/3 of that of the existing algorithm. At the same time the convergence property of particle swarm algorithm is further improved.