Super-Resolution Reconstruction Of Face Image Based On Prior Information

With the advancement of science and technology,images have become one of the main ways for humans to obtain and use information.In daily life,people have higher and higher requirements for image quality,and they are no longer satisfied with unblurred images,but pursue higher-resolution images.Although people continue to pursue higher-resolution images,low-resolution images in real life are often easier to obtain.Current image super-resolution technology can use low-resolution images to obtain high-resolution images.The face image super-resolution in this article belongs to the super-resolution technology of a specific application domain.On its basis,it can perform face attribute recognition,face comparison,face recognition and other facerelated tasks on low-resolution face images.Regarding the face image super-resolution reconstruction,this paper proposes a single-frame face image super-resolution technology based on the geometric features of the face,which uses the prior information of the face.The main research work of this paper has the following two aspects:(1)This paper proposes a progressive residual network based on prior information to achieve super-resolution technology for face images.Firstly,using the prior information of the face,a new method of recovering the face structure through the face analysis map and the coordinates of the key points of the face is proposed,and a super-resolution network structure of the progressive residual is designed.The adopted progressive network is more in line with the network learning process,and can decompose the complex up-sampling task into multiple steps.In the case of a large upsampling amplification factor,it can reduce learning difficulties.The final experimental results show that the super-resolution algorithm in this paper performs better in the quality of the reconstructed image and the texture details of the generated face.(2)On the basis of the foregoing,a face image super-resolution technology based on a generative adversarial network is further proposed.Use the generative confrontation network to improve the details of the generated face image to obtain a better reconstruction effect.This article uses two kinds of generative adversarial networks to compare the effects,and compares the Wasserstein GAN(WGAN)and Relativistic average Standard GAN(Ra GAN)respectively.The experimental results prove that the use of generative adversarial networks is helpful for image reconstruction and restoration.Compared with contrasting super-resolution algorithms,this algorithm has a significant improvement in objective and subjective evaluation.