Research On Image Super-resolution Reconstruction Algorithm Based On Deep Learning

Image super-resolution reconstruction can improve the resolution of the image and restore the high-frequency information of the image.The high-resolution image has more information and the description of the edge detail information is more specific.Nowadays,with the improvement of computer equipment performance and artificial intelligence technology,the application of deep learning to super-resolution reconstruction algorithms has become a research hotspot.The super-resolution reconstruction algorithm based on deep learning uses software to improve the quality of the image,which has the characteristics of low cost and strong flexibility.This paper improves the super-resolution network based on the "depth" of the convolutional neural network,which is mainly reflected in the following aspects:(1)An efficient attention super-resolution convolutional neural network(EASRCNN)is proposed,which is an improved method of the fast super-resolution convolutional neural networks.Firstly,the channel attention module is introduced into EASRCNN to make the model focus on "what features are meaningful".Secondly,the spatial attention module is introduced to divert the model's attention to "where features are meaningful".Thirdly,the residual skip connection is used to combine the local feature information of the front and back networks.Finally,deconvolution is used to reconstruct the image.The experimental results showed that,compared with a basic FSRCNN,the EASRCNN increased the peak signal to noise ratio(PSNR)by up to 0.37 d B and the structural similarity(SSIM)index to 0.9709.(2)A new pyramidal multi-scale residual network(PMSRN)is proposed,which is an improved method of the extremely deep multi-scale residual network.Firstly,the hierarchical residual-like connections and dilation convolution are used to form a multi-scale dilation residual block(MSDRB),the MSDRB enhances the ability to detect context feature.Secondly,the hierarchical feature fusion structure is used to fuse hierarchical local features.Finally,a complementary block of global and local features is added to the reconstruction structure to alleviate the problem that useful original feature information is ignored.The experimental results showed that,compared with a basic multi-scale residual network,the PMSRN increased the PSNR by up to 0.43 d B and the SSIM index to 0.9776.