| In the process of image acquisition and transmission,due to hardware equipment and other reasons,it is easy to be affected by noise,and many details of the image will be lost,making the image vague,resulting in the useful information of the image can not be recognized by people normally,which seriously affects the further analysis and processing of the image.Although the image denoising technology based on convolutional neural network has made remarkable progress and can process the synthetic Gaussian noise image,the existing denoising methods still cannot completely separate the noise from the image.In this paper,two different image denoising network models based on deep learning are proposed for Gaussian noise.(1)Image denoising algorithm of dual-channel extended convolution attention: For image denoising,we proposed CEANet network which has an dual-channel dilated convolution with attention mechanism to solve the problem of information loss caused by deep neural network.Reserving block merged output feature maps of each layer to make up the loss of detail information during convolution.Dilated convolution achieved better balance between denoising performance and efficiency,extracting more features with less parameters and enhancing the representation capability of the model for noisy images.The sparse module of dilated convolution expanded the receptive field to extract significant structural information and edge features and recover details of complicated noisy images.The feature enhancement module based on attention mechanism further guided network for image denoising by fusing global features with local features.The experimental results showed that CEANet achieved high peak signal-to-noise ratio and structure similarity mean value at Gaussian white noise level of25 and 50,which can capture image detail information more efficiently and has better performance in edge retention and noise suppression.Through the above comparative experiments,the effectiveness of the algorithm framework is proved.(2)Image denoising algorithm based on residual-dense block local feature extraction:Aiming at the problem that the existing deep learning denoising algorithm network is too deep,which leads to a large amount of computation,an image denoising network based on local feature extraction of dense residuals is proposed for the removal of Gaussian noise.The local feature extraction module based on extended convolution can expand the receptive field to capture more image details in the convolutional neural network.The input feature map of each layer in the dense block is the output of all previous layers,which effectively uses the feature information,reduces the loss of details in the convolution process,and improves the expression ability of the network.In order to ensure the maximum transmission of feature information extracted from shallow network,the splicing information retention module based on residual connection can train deeper network and ensure good performance.The feature enhancement module based on attention mechanism further guides network denoising by integrating global feature and local feature.The experimental results show that the network achieves good results when the white Gaussian noise level is 25 and 50.The network not only keeps the edge details and texture of the image well,but also captures the image details more efficiently,and also has good performance in noise suppression.The effectiveness of this algorithm for image denoising is proved by the correlation experiment graphs. |
