连续Hopfield神经网络图像复原算法的改进  

Improvement of Continuous Hopfield Neural Network Based Algorithms for Image Restoration

作　　者：吴城磊[1] 

机构地区：[1]南京大学电子科学与工程系,江苏南京210089

出　　处：《电脑知识与技术》2006年第10期99-100,119,共3页Computer Knowledge and Technology

摘　　要：分析了状态连续变化的Hopfield神经网络模型及其收敛性。提出以最大正确转移概率作为更新神经元的选择准则．理论分析表明，该准则能增加神经元状态向提高复原图像信噪比方向转移的概率。利用该准则，提出了串行和并行两种算法进行图像复原．并与连续Hopfield网络复原算法进行比较，发现复原图像的信噪比得到进一步提高。The modified Hopfield neural network model based on continuous state change and its convergence is analyzed, A criterion that has the highest correct transition probability, which the theoretical analysis shows can raise the probability of transition towards increasing the SNR of the restored image, is proposed, A serial algorithm and a parallel algorithm based on the criterion is set up, which prove that, compared with the continuous Hopfield algorithm, the SNR of the restored image can be improved further.

关 键 词：图像复原 连续HOPFIELD网络 最大正确转移概率准则 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]