Research On Image Denoising Algorithm Based On Deep Learning

In recent years,with the popularity of the Internet,data volume has grown rapidly,and hardware capabilities such as that of GPUs have been greatly improved.Neural network and deep learning have gained more and more attention in the technology circle.In 2006,a deep-structured neural network was proposed by Hinton.Since then,artificial neural network has attracted more and more attention.This technology has been successfully applied in various fields,including image and speech recognition,autopilot technology,medical projection and diagnostics,etc.among which image recognition and natural language processing have achieved remarkable results.Deep learning technology is based on a neural network model.Because of its superior nonlinear fitting ability and ability to make the machine capable of simulating human visual and auditory behaviors,it has even exceeded the current limit of human intelligence in some points.Among them,the most familiar to us is AlphaGo which used deep learning technology defeated the Go professional nine-piece player Shishi Li in 2016.In the field of image recognition,the application of deep learning technology has been successful.Similarly,with the continuous development of deep learning technology,outstanding performance has also been achieved in the field of image processing.However,in some areas there are still deficiencies that need continuous improvement and improvement.This paper is based on the application of deep learning techniques for image denoising.In real life,because of imperfect equipment and systems,images are often blurred by noise.In order to make the blurred noise image clearer and the image features more obvious,this article uses neural network to achieve image noise reduction with good learning ability of statistical characteristics of the image.Mainly based on Convolutional Neural Network(CNN)algorithm,the effect of activation function on network optimization is mainly studied.In the deep network,multiple feature extraction techniques are used to learn the more abundant features of the input image.The inverse propagation of convolutional neural network is optimized to accelerate the training speed of the model and improve the convergence of the algorithm by a better adaptive algorithm.Combined with batch standardization and residual learning techniques,an image denoising network with better noise reduction performance is designed based on the image denoising model with deep residual learning and CNN(DnCNN).Finally,a comparison is made between the algorithm mentioned this paper with other excellent denoising algorithms.From the results,it can be seen that the denoising algorithm improved in this paper can also improve the detailed recovery of denoised images without loss of sharpness.And under different noise standard deviations,better peak signal-to-noise ratio(PSNR)is obtained than other outstanding denoising algorithm.It is proved that the improved convolutional neural network noise reduction model in this paper is very competitive.