Research On Deep Learning Based Image Super-Resolution Algorithm

Image is one of the most important sources of information for human beings.However,due to the quality problems of the imaging device and the acquisition environment,image quality is damaged and the resolution is reduced.To solve this problem,the image super-resolution algorithm can reconstruct the degraded image to improve the image quality and visual effect.Recently,deep learning has become a research hotspot in the field of image super-resolution.This thesis focuses on the research of image super-resolution algorithm based on deep learning,and the main research contents and achievements are as follows.To solve the problem of remote sensing image super-resolution,a non-locally up-down convolutional attention network is proposed.First,a non-local features enhancement module(NLEB)is constructed to obtain the spatial context information of high-dimensional feature maps,which allows our network to utilize global information to enhance low-level visual features with effect,overcoming the defects of deficiency perceptual ability of shallow convolutional layers.Second,an enhanced up-sampling channel-wise attention(EUCA)module and enhanced down-sampling spatial-wise attention(EDSA)module are proposed to weight the features at multiple scales.By integrating the channel-wise and multi-scale spatial information,the attention modules can compute the response values from the multi-scale regions of each neuron and then establish the accurate mapping from low to high resolution space.Extensive experiments on NWPU-RESISC45 and UCMerced-LandUse datasets show that the proposed method can provide state-of-the-art or even better performance in both quantitative and qualitative measurements.To solve the problem of face image super-resolution,a deep end-to-end face super-resolution reconstruction algorithm based on multitask joint learning method is proposed.The model uses residual learning and symmetrical cross-layer connection to extract multilevel features and adopts the encoder-decoder structure.Then,a joint training method is designed for face multi-attribute learning tasks:setting loss weights and loss thresholds on the basis of the learning difficulty of different tasks and avoiding the influence of subtasks on the learning of head tasks after fitting the training set to obtain considerable abundant face prior knowledge information.Experimental results on CelebA show that the proposed algorithm can further utilize the face prior knowledge and create further realistic and clear facial edges and texture details in visual perception.To solve the problem of natural scene image super-resolution,a novel Multi-Granularity Pyramid Attention Network is proposed,which fully exploits the multi-granularity perception and attention mechanisms to improve the quality of reconstructed images.A multi-branch dilated convolution layer with varied kernels corresponding to receptive fields of different sizes is constructed to modulate multi-granularity features for adaptively capturing more important information.Moreover,a novel spatial pyramid pooling attention(SPPA)module is constructed to integrate the channel-wise and multi-scale spatial information,which is beneficial to compute the response values from the multi-scale regions of each neuron,and then establish the accurate mapping from low to high dimensional solution space.Besides,for long-short-term information preservation and information flow enhancement,we adopt the short,long,and global skip connection structures to concatenate and fuse the states of each module,which could improve the SR network performance effectively.Extensive experiments on several standard benchmark datasets show that the proposed method can provide state-of-the-art or even better performance in both quantitative and qualitative measurements.