摘要
提出了用遗传算法优化的神经网络和PSO算法优化的神经网络图像复原算法,并对它们的复原效果进行了比较.先用优化算法对神经网络的结构进行优化,然后利用优化后的神经网络的学习和泛化能力,用一组样本图像对网络进行训练,建立退化图像与原图像之间的非线性映射关系,最后用训练好的进化神经网络对待复原的退化图像进行图像复原.实验结果表明复原的图像无论在主观视觉还是定量分析上都取得了很好的效果.
Image restoration method based on neural network combines with GA and image restoration method based on neural network combines with PSO are proposed, and their effects of image restoration are compared too. First, the structure of the neural network are optimized by optimization algorithms. Second,nonlinear mapping relationships between the blurred image and clear image are established by training the optimized neural network which has the ability of learning and generalizing with a group of sample image. At Last, blurred image which needs restoring could be restored by the trained evolving neural network. The results of experiments demonstrate that this method has a satisfying restoration effect both in visual impression and quantitative analysis.
出处
《四川大学学报(自然科学版)》
CAS
CSCD
北大核心
2011年第5期1081-1086,共6页
Journal of Sichuan University(Natural Science Edition)
关键词
图像复原
神经网络
遗传算法
PSO算法
image restoration, neural network, genetic algorithm, PSO algorithm