Edge And Attention Assisted Single Image Super-resolution

Image super-resolution methods have been trying to obtain a high-resolution image from a low-resolution image.Various methods have been presented in this vein by using multiple techniques.Learning-based methods have been struggling to get higher perfor-mance for a lot time and then after arrival of deep learning,convolutional neural network has revolutionized every field of computer vision,including single image super-resolution because of its remarkable performance pertaining to effectiveness and efficiency.Among other CNNs architecture,residual network architecture is being used a lot in recent methods because of its deeper approach and higher accuracy.On the other hand,few of the SR meth-ods try to predict the SR image by incorporating different prior knowledge of images.These prior can be neighbor embedding,sparse prior or edge prior.Those methods who use edge prior in their method,have been using a single edge map to obtain better SR performance.But these methods still lack the sharper edges in output.In this paper,we propose a new method that utilizes not only internal image features but also multi-level edge prior knowledge with richer information.Holding the intuition that edge information helps to deal with blurry edges and try to generate sharper results,we present a residual edge and channel attention super-resolution network to handle LR images,named RECAN.Our architecture consists of two basic modules:the first module is EdgeNet,which generates multi-level edge maps from the input image;and the second module takes advantage of significant information in input image along with edge maps,called SRNet.Specifically,the SRNet uses channel attention technique and spatial feature transform(SFT)layers to super-resolve an image.We train and test our model using benchmark datasets to show its effectiveness.Qualitative and quantitative comparisons are presented with state-of-the-art methods for the scaling factor of ×2,×4 and ×8.These comparisons show the promising results of our method together with improved image quality.