Image Super-resolution Reconstruction Based On Deep Residual Network

With the rapid development of the information age,people have higher and higher requirements for the quality of the image.The higher the resolution of the image,the richer the information in the image,which is conducive to people to obtain more comprehensive information.Due to the influence of realistic factors such as acquisition equipment and environment,the resolution of acquired images is not good,and the cost of improving image resolution by using hardware equipment is high,which is not conducive to operation.The super resolution reconstruction technology realized by software not only saves cost but also has better effect,which makes its application more and more widely.Aiming at the problem that the existing image super-resolution reconstruction algorithm is not comprehensive in extracting image details,this paper takes this as the starting point to carry out research,and the main work is as follows:(1)In view of the traditional algorithm to extract details of poor ability,information can not be fully used,this paper proposes a deep residual network based on FRactional Fourier Transform(FRactional Fourier Transform,FRFT)image super resolution reconstruction algorithm,Firstly,the corresponding low-resolution image is obtained by using the quality reduction model,and the shallow features of the image are extracted by using the convolutional layer.Secondly,the introduction of residual blocks speeds up network convergence,and FRFT is added to the residual block to transfer the image to the fractional domain for discretization calculation,and then back to the airspace,so that improve the ability of the network to extract frequency detail information such as texture.Finally,deconvolution layer is used to realize the upsampling of the image,and the low-resolution image input from the network is fused with the image sampled from the network to realize the image reconstruction.(2)In order to alleviate the problem that the image features extracted by the convolution kernel at a single scale are not sufficient,which may easily cause the phenomenon of information insufficiency and consumption,this paper proposes an image super-resolution reconstruction algorithm based on a multi-scale dense residual attention network.The structure of the network is mainly consists of three parts,the first part is multi-scale feature extraction network,the middle structure is the dense residual attention module,and finally subpixel convolutional network.At the beginning,shallow feature extraction is performed on the input image using convolution kernels of different sizes,and the obtained results are fed into the next part of the network.Then intensive residual attention module is used to extract the features of the image,so that the feature information is more comprehensive.Finally,subpixel convolution is used to complete the up-sampling process,and the final reconstructed image is obtained.(3)Experimental results and analysis.After the proposed algorithm model,the effective-ness of the proposed algorithm is verified by specific experiments.The first algorithm mainly analyzes the optimal order and the number of residual blocks of FRFT,and compares the results of the proposed algorithm with those of other algorithms.It proves that the proposed algorithm has better reconstruction effect and better evaluation index result value.In algorithm 2,multi-scale and single scale are added for feature extraction,and the effectiveness of adding dense connection and attention mechanism is analyzed.The experimental results are compared with other algorithms,and it is proved that the proposed algorithm achieves better results and gets higher evaluation index value.