Research On Image Super-resolution Reconstruction And Image Inpainting Based On Deep Learning

With the development of the times and the progress of society,the ways in which people obtain information have become more diversified and simplified.As the main carrier of information,images are still the main object for people to obtain information.However,due to the interference of various factors,there is a gap between the obtained image and people's expected image.Coupled with the improvement of quality of life or work needs,ordinary images can no longer meet the needs,the image reconstruction and image restoration technology can just optimize the visual quality of the image,so as to fully exploit and utilize the information of the image.Therefore,image reconstruction and repair technology has become one of the most popular research directions in the field of computer vision.It is widely used in photo modification in daily life,image recognition in security work,and reconstruction and repair of cultural relic photos.In this context,this article takes image super-resolution reconstruction and image restoration as the main research content,and conducts in-depth study and research.In the study of super-resolution reconstruction and image restoration,the following algorithms are proposed to solve the problems of excessive enhancement of edge information,insufficient detail recovery ability,and excessive recovery in the existing processing technologies:(1)For the disadvantages of traditional SRCNN(Super resolution using convolution neural network,SRCNN)algorithm,incomplete image feature extraction and insufficient information utilization,an improved multi-branch reconstruction algorithm is proposed.First,for the sake of more fully extract image feature information,the network depth is deepened to 4 layers.Secondly,multi-channel object reconstruction is performed by setting convolution kernels of different scales,so that more information can be considered in the reconstruction process.Finally,the PReLU function is used in the network structure to solve the problems of neuron necrosis and difficulty of convergence.Experiments show that the improved algorithm achieves a better reconstruction effect,and the peak signal-to-noise ratio and structural similarity are better than the traditional SRCNN algorithm.(2)Aiming at the disadvantages of the traditional Anchored Neighborhood Regression(ANR)image super-resolution method that is inflexible and has no good recovery ability to the details of the image,an image reconstruction method that combines Anchored Neighborhood Regression and Convolution Neural Network(CNN)is proposed.Firstly,the elastic network regression model is proposed in ANR to make the algorithm have the characteristics of feature selection.Secondly,lanczos3 interpolation method is used in the image preprocessing part of CNN to speed up the operation speed.In feature extraction,Swish function with self-gating feature is proposed as an activation function to improve the test accuracy.Finally,the correlation coefficients of reconstructed images are proposed to evaluate the validity of reconstructed images.The simulation results prove that the algorithm effectively constructs the details of the image,and the image quality has been significantly improved.(3)Aiming at the shortcomings of the traditional Criminisi algorithm which the priority value tends to zero quickly and costs much inpainting time,an improved image inpainting algorithm is proposed based on information entropy and gradient factor.First,the information entropy and the gradient factor for the image are fitted as weight factors,and the priority calculation method is optimized to find the optimal inpainting block.Second,the information entropy which can measure the complexity of the pixel block is used to adjust the search area of the matching block to establish a dynamic rule of the search area.Then,an adaptive model of the template size for the matching block is established with the help of the gradient factor to improve the optimal matching block search strategy.Finally,the sequential similarity detection algorithm(SSDA)is introduced to select the optimal matching block from the source region to achieve image inpainting.The simulation results prove that,compared with the traditional Criminisi algorithm,the proposed algorithm has obtained satisfactory repair results in both objective and subjective aspects,the recovery effect is more real and natural,and the visual effect is better.