基于PSO_Trainlm BP模型的图像去噪研究

针对在使用BP模型进行图像去噪时,模型存在的对初始权阈值敏感、易陷入局部极小值和收敛速度慢的问题.为了提高模型去噪效率,提出采用改进粒子群神经网络模型进行图像去噪.首先运用改进粒子群算法对BP神经网络权阈值进行初始寻优,再用trainlm BP算法对优化的网络权阈值进一步精确优化,随后建立基于粒子群算法的BP神经网络去噪模型,并将其应用到图像去噪研究中.仿真结果表明,新模型结合了粒子群算法的全局寻优能力和BP算法的局部搜索能力,减小了模型对初始权阈值的敏感性,有效防止了模型陷入局部极小值的可能,提高了图像去噪模型的速度和质量.

Researching Image Denoising Model Based PSO_ Trainlm BP

For the denoising operation with the BP model,the model sensitivity to initial weight threshold,easy to fall into local minima and slow convergence.In order to improve the model denoising efficiency,improved particle swarm neural network model is proposed to use denoising operation.First,the use of improved particle swarm algorithm instead of the initial BP neural network optimization,Further precise optimize optimize network weights threshold then the trainlm BP algorithm.Followed by the establishment of denoising model of BP neural network based on particle swarm algorithm and its application to the study of image denoising.The simulation results show that the new model combines local search advantage of the global optimization capability of particle swarm algorithm and BP neural network algorithm,reduced the sensitivity of the model on the initial weight threshold,effectively prevent the network may fall into local minimum value to improve the speed and quality of the image denoising model.