Improvement of Continuous Hopfield Neural Network Based Algorithms for Image Restoration
WU Chenglei, Nanjing · Computer Knowledge and Technology · 2006
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.