Image Denoising Algorithms Based On Convolution Neural Network And Wavelet Transform

Due to various factors such as imaging equipment,images will be interfered by noise in the process of imaging or sensing.Image denoising aims to reduce or eliminate the impact of noises on the image,which often leads to the loss of high-frequency information.Noises will seriously reduce the visual quality of the acquired image,which result in the decline of the reliability of image information,and affect further image processing.Noise removal is an essential step in various image processing and computer vision work.The quality of its processing results directly affects the smooth progress of various follow-up tasks such as edge detection,target recognition,etc.Therefore,it is of great significance to study and optimize image denoising algorithms.Deep learning model is based on probability statistics and applied mathematics.Thanks to the powerful computing power of computer,relevant researchers have made breakthroughs in large-scale data sets,and deeper artificial network model has been established.In the fields of computer vision,speech recognition and natural language processing,artificial neural network has made some achievements.In recent years,it has gradually become the focus of big data and artificial intelligence on the Internet.The image processing algorithm based on convolutional neural network is developing continuously,which provides a new idea for image denoising.In this paper,convolution neural network and wavelet transform are used in image denoising and three image denoising algorithms are proposed.Firstly,a convolutional neural network with information reservation block is designed,which has relatively low computational complexity.The information reservation block extracts the mixed feature information of local long path and local short path through residual learning.The algorithm has good effects on different intensity of Gaussian white noise,Poisson noise and salt and pepper noise.Next,combining with convolution neural network and wavelet transform,an effective algorithm for natural noise and noise in remote sensing image is proposed in this paper.First the image is decomposed by wavelets by the algorithm,and then the different components are inputs to the network for training.Finally,Based on the WNNM denoising algorithm,an improved WNNM image denoising algorithm is provided in this paper.The image processed by WNNM is processed by a convolutional neural network with multi-path information fusion module to enhance the texture details of the image.Among the three algorithms mentioned above,the evaluation indexes used in this paper are SSIM,PSNR,BRISQUE and NIQE,which are the reference evaluation indexes to objectively quantify the experimental results.The experimental results show that the three algorithms proposed in this paper are superior to the image denoising algorithms compared in this paper in terms of subjective visual effect and objective quantitative indexes.