Research On Face Super-Resolution Reconstruction Technology Based On Prior Information

Face super-resolution reconstruction technology is a technology to reconstruct low-definition face images into high-resolution face images.Face super-resolution reconstruction technology is widely used in the fields of face recognition,face detection and facial expression analysis,and has important research value and significance.After the application of deep learning technology in the field of face super-resolution,face super-resolution technology has developed rapidly.However,when the super-resolution factor is large,it is still difficult to recover high-frequency information such as texture and edge in face image.The existing face super-resolution methods using prior information can improve the reconstruction effect of face image.Therefore,based on the prior information of face analysis image and face key points,in order to solve the problem that it is difficult to recover the edge information of face image,this paper introduces the face edge information into the network,and in order to improve the texture recovery effect,the texture loss function is introduced into the loss function of the network,The network also introduces the attention mechanism to improve the ability of the network to use the feature information.The specific research contents are as follows:1)A face super-resolution reconstruction method based on face edge and analytical graph is proposed.When the super-resolution factor is large,it is difficult for the face super-resolution method to recover the high-frequency information such as face edge and texture.The face edge information and face local analytical graph are introduced into the face super-resolution reconstruction network as a priori constraints,the texture loss is introduced into the loss function of the network,and the attention mechanism is introduced into the network.Through experiments,it can be found that under the super-resolution factor of 8 times,the network can reconstruct a clearer face image.Moreover,by adding face edge information,texture loss and attention mechanism,the reconstruction effect of the network has been significantly improved.2)A face super-resolution reconstruction method based on face edges and key points is proposed.Aiming at the problem that it is difficult to recover the face edge information and texture information,the network takes the low-definition image and its edge information as the input.The loss function of the network also introduces the texture loss,and the network also introduces the attention mechanism.Aiming at the problem that the prior information extracted directly from the low-definition image may be inaccurate,an iterative cyclic network is designed to reconstruct the high-resolution image by extracting face key points from the reconstructed image and generating new feature information.Through experiments,it is found that adding face edge information,texture loss and attention mechanism under 8 times super-resolution factor improves the reconstruction effect of the network,and the reconstruction effect of the network is better than that of the network in work 1.