Research On Single Image Super-Resolution Reconstruction Method Based On Fusion Of Internal And External Features

Image Super-Resolution(SR)reconstruction technology is to reconstruct High-Resolution(HR)images from observed Low-Resolution(LR)images.It is an important technique in the field of image processing.This technology has very important practical application value in many fields such as surveillance,satellite,and medical imaging.In recent years,with the rapid development of deep learning technology,Convolutional Neural Network has made remarkable achievements in Single Image Super-Resolution(SISR)technology,in which the residual network shows better performance.Based on the research of the single image super-resolution reconstruction algorithm based on deep convolutional neural network,this paper studies how to effectively use the internal feature information of the image itself to overcome the problems of the lack of details of the high-resolution reconstruction image caused by the low utilization of the internal feature information of the image.In order to make full use of the internal feature information of images,a full convolutional deep neural network is proposed to extract the internal features of LR images.Meanwhile,this paper proposes a Super-Resolution Fusing Internal and External Features Network(SRIFN).This network uses the internal features of the image itself as the prior knowledge,then,the extracted internal features are embedded into the residual network using the Spatial Feature Transform(SFT)module to form a new residual module,so as to establish a deep residual network based on the fusion of internal and external features.And further proposed a single image Super-Resolution based on Progressive Upsampling Fusion(SRPUF),the post-upsampling SISR network based on the fusion of internal and external features was changed to a progressive upsampling SISR network.At the same time,the algorithm uses deconvolution and pixel reconstruction to perform pixel fusion of the two up-sampled HR images to improve the reconstruction effect.It has strong early reconstruction capabilities,can gradually generate the final high-resolution images,and further improve the performance of the single-image super-resolution reconstruction network based on the fusion of internal and external features.The experimental results show that the method proposed in this paper enhances the quality and texture characteristics of the reconstructed image,and improves the visual effect of the reconstructed image.