Deep Learning Image Restoration Algorithm Based On Small Samples

Image restoration has always been a hot topic in the field of computer vision.In recent years,with the rapid development of artificial intelligence,deep learning has also developed rapidly.In particular,there has been tremendous development in the fields of image recognition,image classification,speech recognition,identification,target tracking,image restoration,and behavior analysis.The deep learning of the deep problem learning and the rapid and effective extraction of physical features such as images and speech make these problems that need to be manually set and manually extracted in the traditional method.Deep learning model training and learning becomes more convenient and faster.Although deep learning effectively solves the model fitting of some nonlinear problems,the implementation of the deep learning model requires a large amount of training data.In the field of image restoration,the traditional method can achieve image restoration by calculating the information around the cavity in the case of small hole coverage.However,in the case of large voids,the traditional method cannot achieve effective information association because the cavity area is too large,so that the information in the cavity cannot be accurately recovered.Deep learning can achieve semantic recovery of the void region by extracting advanced features from the image and achieving semantic recovery.The existing deep learning image repair algorithms are often based on the large sample data,and the small sample data often does not have a good repair effect.Aiming at the small sample situation,this paper proposes an algorithm against circular convolutional neural network,which makes full use of the cyclic network to extract multi-level image features.The repair of images of small sample data in the case of a small number of network layers is achieved.The experimental results show that a better result is obtained.The main contributions of this article are:(1)Summarize the development status and research status of image restoration technology,then introduce the main problems faced by image restoration based on deep learning,and give a brief description of the article chapter arrangement.(2)An algorithm against circular convolution network is proposed.The network is divided into two parts: the generation network and the decision network.The generated network part is a circular convolution network,which realizes image repair in a small sample case.The decision network part is a twin network structure,which enhances the quality of generating network picture generation and makes the network as a whole more robust.The comparison experiments show that the proposed algorithm has a good effect on image restoration.(3)A brief summary of the work of the thesis,and a prospect for future work and innovation.