Application Of Improved Convolutional Neural Network In Images Super Resolution

Image super resolution is a kind of technology that uses network model to input the blurred image,and then carry out special model processing,and finally output the clear and high-quality image.Due to the shortcomings of the traditional super-resolution network model in the processing of low-resolution images,such as slow convergence speed in training,more dependence on the image region itself and poor visual effect of the reconstructed image,etc.A series of researches and improvements have been made to improve the super-resolution technique.This is an article study to improve the convolutional neural network application in the image super-resolution problem.The network is in the traditional network on the basis of super-resolution respectively to the traditional model framework,the sampling method of super-resolution network and network design three aspects to improve.And in view of the existing image super-resolution algorithm based on convolution neural network parameters,large amount of calculation,more training time is longer,fuzzy image texture and other issues to the existing image classification Network model and visual recognition algorithm,an improved Generative Adversarial Network is proposed.The Model derives better output through the interactive learning of the modules in the framework,namely,the Generative Model and the Discriminative Model.In addition,through the comparison between the reconstructed image effect of the traditional super resolution convolutional neural network and the improved super-resolution convolutional neural network,it is concluded that the improved super-resolution network model has better visual effect,higher PSNR value and stronger image resolution than the traditional one.