摘要
研究图像提取问题,在处理由不同种类纹理区域组成的彩色图像时,针对克服量子粒子群优化(QPSO)聚类算法由于早熟现象导致图像分割过程中难以计算出精确纹理区域,为了能准确提取图像目标和提高精度,提出了基于高斯扰动的量子粒子群优化(GQPSO)的新型聚类算法。受益于高斯扰动,GQPSO改善了QPSO固有的多样性下降和陷入局部早熟的问题,而快速逼近全局最优解。对Berkeley Segmentation数据库中的6幅图像的分割实验结果表明,相比于PSO和QPSO,GQP-SO的聚类效果和性能均有明显改善。
To overcome the shortcomings that Quantum-behaved Particle Swarm Optimization（QPSO） is difficult to compute the exact texture fields for the color images composed of different kinds of texture regions,due to its premature phenomena,in this paper,a new algorithm named Gaussian disturbance based Quantum-behaved Particle Swarm Optimization（GQPSO） is proposed.Benefiting from the Gaussian disturbance,GQPSO can fast approximate the global optimal solution after solving the problems of QPSO with inherent diversity declining and partially leading to minimum.The simulation results of six images from Berkeley Segmentation dataset have demonstrated that the GQPSO clustering algorithm is more efficient and perfect than the existing PSO and QPSO.
出处
《计算机仿真》
CSCD
北大核心
2011年第3期275-278,共4页
Computer Simulation
基金
国家自然科学基金(6077320660704047
60572034
90820002)
教育部新世纪优秀人才计划(NCET-06-0487)
江苏省自然科学基金(BK2006058)
关键词
量子粒子群优化算法
高斯扰动
聚类
图像分割
Quantum-behaved particle swarm optimization（QPSO）
Gaussian disturbance
Clustering
Image segmentation
