Research On Face Image Super Resolution Reconstruction

With the development of technology and the increasing awareness of social security,face recognition system has been widely used in various fields of public safety.However,in many real scenes,most of the captured images are low resolution face images,which are difficult to be used because of the loss of high frequency information.Image super-resolution reconstruction algorithm can effectively solve this problem,which can reconstruct the high resolution image from the low resolution image.Human face is a very important target in video surveillance.Therefore,the face image super-resolution reconstruction algorithm is a popular research direction in recent years.The human face image has a very good structure(the face is composed of facial features,and the relative position between the facial features is fixed),and this structure has a universal similarity(the facial structure of all people in the world is very similar,regardless of gender,race and skin color).For face image super-resolution reconstruction,we believe that the importance of different regions of the face image is different,so we should treat them differently.Based on the above assumptions,this paper proposes two kinds of weighted face image super resolution reconstruction algorithms.One is based on information entropy,and the other is based on the Adaboost algorithm to update the weights.The experimental results show that the proposed two weighted face image super-resolution reconstruction algorithm can effectively improve the quality of the generated high resolution face image.In addition,in the actual application scene,the face image captured by camera not only is low resolution,but also there are a variety of changes on their faces,such as facial expressions,lighting,occlusion,etc.For this kind of low resolution face images with various facial changes,this paper proposes a new algorithm for the removal of various facial changes.The algorithm is divided into two steps:the first step is to transform a low resolution face image with a variety of facial changes into a low resolution face image without face changes;the second step is based on the first step,the high resolution face image is reconstructed from the low resolution face image with no face changes.Through these two steps,we can reconstruct the low resolution face image with various facial changes into a high resolution face image without face changes.The experimental results show that our proposed algorithm is effective for the removal of various facial changes.