The Research On Image Super-resolution Reconstruction Technology

Images are the main medium for humans to get information from the outside world.But influenced by the system and hardware conditions,the degradation of image always occurs during the process of acquisition,compression,storage,and transmission.The degradation including low spatial resolution,low contrast,blurred edges,detail lost,and noise interference introduced in the process of information exchange,unavoidably,affect the adequacy and accuracy of image information expression in image super resolution.So,improving image resolution has become a hot issue in the current image processing field.To rich information of the image,we use super resolution(SR)algorithm through software algorithms to enhance image resolution.The resolution enhancement means that the image has more useful details which can help with other visual tasks.Therefore,SR is widely used in video surveillance,target detection,face recognition,automatic driving,medical diagnosis,etc.To settle the problem of insufficient detail information and limited numbers of samples in the existing image resolution reconstruction,this paper studies networks including Laplacian pyramid network,zero-sample learning(ZSL)network and attention-based generative adversarial network.In this paper,we first introduce sub-pixel convolution and guided filter to propose a visible light-guided infrared image super-resolution reconstruction algorithm.Second,we use dilated convolution and bottleneck network to construct a zero-sample superresolution model.In the last,we take self-attention and channel attention mechanisms into an attention mechanism based super-resolution network.The main researches are summarized as follows.(1)To solve the problem of insufficient detail and excessive noise in infrared images.The RGB-IR cross input and sub-pixel up-sampling network is proposed to increase the spatial resolution of an Infrared(IR)image by combining it with a color image of higher spatial resolution obtained with a different imaging modality.Specifically,this is accomplished by fusion of the features map of two RGB-IR inputs in the reconstruction of an infrared image.To improve the accuracy of feature extraction,deconvolution is replaced by sub-pixel convolution to up-sample image in the network.Then,the guided filter layer is introduced for image denoising of IR images,and it can preserve the image detail.In addition,the experimental dataset,which is collected by us,contains large numbers of RGB images and corresponding IR images with the same scene.(2)Based on a small number of training samples,this paper proposes a novel zero-shot learning(ZSL)network for super resolution to learn internal self-similarity in the test image.In this model,the training samples are cropped from the input image,which means no extra dataset is required for model training,the overfitting problem in this way can be well avoided.Besides,since all the training samples are cropped from the input itself,the nonlocal selfsimilarity attributes of the test image can be fully utilized.Moreover,the efficient spatial pyramid of dilated convolutions network with a bottleneck(ESP-BNet)is applied in the model as an efficient computational structure to enhance the feature representation.(3)To exploit more channel detail of panchromatic images,a self-attention and channel attention mechanisms are introduced in a Wasserstein adversarial network based on selfattention enhancement for panchromatic image super-resolution,which can obtain hidden feature information from multiple dimensions of channels and spaces.In this model,we use an encoder-decoder network followed by a fully convolutional network(FCN)as the backbone to extract multi-scale features and reconstruct the High-resolution(HR)results.To exploit the relevance between multi-layer feature maps,we first integrate a convolutional block attention module(CBAM)into each skip-connection of the encoder-decoder subnet,generating weighted maps to enhance both channel-wise and spatial-wise feature representation automatically.Besides,considering that the HR results and Lowresolution(LR)inputs are highly similar in structure,yet cannot be fully reflected in traditional attention mechanism,we therefore design a self augmented attention(SAA)module,where the attention weights are produced dynamically via a similarity function between hidden features,this design allows the network to flexibly adjust the fraction relevance among multi-layer features and keep the long-range inter information,which is helpful to preserve details.In addition,the pixel-wise loss is combined with perceptual and gradient loss to achieve comprehensive supervision.The above algorithms are verified by rigorous experiments.The results of different superresolution scales on the benchmark data set show that the proposed method in this paper is better than the existing traditional algorithms and most deep learning algorithms in the image super-resolution reconstruction.