改进的粒子群算法用于图像分割

Image segmentation with Otsu based on improved PSO

提出了一种将保证收敛粒子群算法与最大类间方差法相结合的快速阈值分割法。该方法根据最大类间方差法的原理以分离度大小作为判断粒子优劣的准则,即分离度越大粒子就越好,并采用粒子群算法对图像进行多目标优化搜索。实验表明,该算法在继承标准粒子群算法易于实现、实时性好等优点的同时,还避免了标准PSO算法存在的早熟收敛问题,具有更强的寻优能力。

Threshold method is one of the most basic and important techniques for image segmentation because of its simple-realization and good result. This paper proposes a rapid threshold method which combines the GCPSO with Otsu. This method uses the separation level to judge whether a particle is good or bad based on Otsu principle, and the GCPSO is designed to optimize the decision variables. The experiment shows that this algorithm retains the simple-operation and good real time of standard PSO, and also solves the premature convergence problem, thus it has better search capacity.