Research On Face Image Replacement

Face replacement is an example of image recognition and application.It USES image processing technology to replace the face information in the image,so as to achieve the effect similar to face changing.Face replacement technology plays an important role in film and television creation,game design,image processing and criminal investigation.This article through to face to replace the technical part of the technology is analyzed,and involved in the process of implementation of face replacement algorithm of face recognition,feature extraction,feature fusion,color replacement and face replacement technology research,comparison of face replacement effect,deep learning algorithm is applied to the replacement of face problems existing in the process of optimization,and complete to replace the face in the static image automatically.The application of deep neural network in image processing is a hot spot at present,which is used to process more complex information.In view of the shortcomings of traditional face replacement methods,this paper proposes a face synthesis method based on multi-layer deep convolutional network.The main research contents and innovation points are as follows:the traditional face replacement technology in the treatment of the face replacement method,the low accuracy,authenticity,for facial features detection and extraction effect is not very good,and in the face after the fusion effect is not very ideal,at the same time in the treatment of the face facial shade,illumination is not clear,the influence of such factors as lead to face replacement effect is not very good.This paper will combine the classic "three courts and five eyes" model to extract facial features,facial posture and other data,and form a feature database,which will be used as the data set of the depth prediction model,and build a depth convolution model to analyze the features;this paper proposes a face replacement algorithm based on the two-layer neural network model,which is used to predict and process the human features and posture in each link.The first layer is used to train the facial features in the image,and to align the face or estimate the posture.The second layer compares and fuses the data of the first layer to obtain a face replacement image with higher accuracy and authenticity.The third layer of neural network is used to estimate the head pose in order to better achieve face replacement;compare the face replacement algorithm studied in this paper with the traditional active appearance model,poisson image fusion,self-coding image replacement and other algorithms,and analyze the feasibility of this algorithm and its effect in image replacement processing.Through experimental comparison,the algorithm proposed in this paper has higher accuracy,stronger sense of reality and easier implementation in face replacement,with good face replacement effect and high robustness.