双群交换微粒群算法及在Shearlet图像去噪中的应用

Two sub-swarms exchange particle swarm optimization and its application in Shearlet image denoising

分析基于不同进化模型的双群交换微粒群优化算法的不足,提出改进的双群交换微粒群优化算法。算法将微粒分成大小相同的两分群,第一分群采用标准微粒群模型进化,第二分群采用Cognition Only模型进化,当微粒进化到稳定状态,从第一分群随机抽取部分粒子与第二分群适应值最差粒子进行交换,重复上述操作直到找到最优解。实验结果显示:该算法有更好的全局寻优能力和达优率。为验证算法实用性,将改进算法用于Shearlet图像去噪。该方法根据Shearlet变换域不同尺度和方向系数的分布特性,采用改进算法自适应确定各尺度和方向的最优阈值,实现基于图像内容的自适应去噪。实验表明,该方法能有效滤除图像噪声,较好保留图像边缘信息,去噪后图像具有更高峰值信噪比(PSNR)。

By analyzing the deficiencies of two sub-swarms exchange particle swarm optimization based on different evolutionary models,an improved two sub-swarms exchange particle swarm optimization algorithm is proposed.The algorithm divides the particles into two sub-swarms of the same size,the first sub-swarm using the standard PSO evolutionary model and the second sub-swarm using the Cognition Only evolutionary model.When the particles evolve to a stable state,part of the particles from the first sub-swarm are randomly selected to exchange with particles with the worst fitness values from the second sub-swarm.The above-mentioned operations are repeated until the optimal solution is found.Experiment results show that the proposed algorithm has better global search capability and optimal convergence rate.To verify the practicality of the algorithm,the improved algorithm is applied to Shearlet image denoising.According to the distribution characteristics of Shearlet transform domain coefficients of different scales and directions,the method uses the improved algorithm to adap- tively determine the optimal thresholds of different scales and directions,to achieve image content-based adaptive denoising. Experiment results show that the method can effectively filter out image noise and better retain edge information,and the denoised images have higher Peak Signal to Noise Ratio（PSNR）.