摘要
针对图像复原提出了神经元状态连续变化的Hopfield神经网络模型,详细讨论了两种连续函数串行、全并 行复原算法的收敛性和参数选择,仿真实验表明,该模型能够精确达到能量极小点,并对复原图像的信噪比有一定的提高。
A modified Hopfield neural network model based on continuous state change is proposed to restore a degraded image. In this paper, we discuss the serial and the parallel algorithm of two continuous functions respectively, and thoroughly study the convergence and the choice of parameters. Experimental results demonstrate that this model can obtain the minimum of the energy and the SNR of the restored image has some improvement.
出处
《信号处理》
CSCD
2004年第5期431-435,共5页
Journal of Signal Processing
