Face Synthesis Based On Triangular Mesh Warp

In recent years,with the advent and development of artificial intelligence era,largescale human-computer interaction technology has attracted more and more people’s attention,as a key part of human-computer interaction technology,image processing technology has become a hot research direction.Image processing contains a lot of research fields,among them,face synthesis technology is an important branch of image processing.By controlling the features of the input face images and using the image deformation algorithm,the input face images are deformed smoothly,and a natural and realistic synthetic image is obtained by linear combination of the deformed face images.The synthetic face has the features of each input face at the same time.Synthetic virtual face has great application value in commercial entertainment,human-computer interaction,data enhancement and other fields.This paper mainly studies the method of face image synthesis,and proposes an algorithm framework that can automatically synthesize two or more faces,moreover,each algorithm module in the framework is independent and easy to be replaced or expanded.The specific research contents and results are as follows:1.This paper uses the Ensemble of Regression Tree algorithm to extract face feature points,the feature points of the target face image are obtained by linear combination of the feature points of the input face images,then,the delaunay triangulation is carried out with each set of feature points,and then,according to target triangles,original triangles are deformed,the target face image is obtained by linear combination of the deformed face images,finally,the synthesized face image is superimposed on the designated area of the background image to obtain a complete fusion image.2.This paper aims to solve the problem of artificial tooth synthesis in face image.The traditional synthesis method is to directly deform the entire face region,and the face expression is not taken into account in the deformation process,it may cause the tooth area of the target image to be composed of tooth texture and skin texture,and the final result is unnatural.In this paper,the tooth region is proposed to be synthesized separately,it is suitable for the case that the tooth area of input face images have different degree of extension and closure.This method improves the traditional face synthesis,and can synthesize a face image of different facial expressions naturally and realistically.3.This paper solves the problem of different color in face synthesis.When the synthetic face image is superimposed on the background,due to images taken under different lighting conditions and equipment have very different colors,the final result is likely to be an odd color which makes person feel unnatural visually.In this paper,poisson editing method is used to adjust the color of face images by using the gradient of synthetic face image and the fusion boundary of background image,and it improves the synthesis very well.