Face Image Beautification And 3D Reconstruction Application Based On Feature Fusion

Three-dimensional face reconstruction is an important research content in computer graphics and computer vision.It has a wide range of applications in three-dimensional animation,video games,virtual clothing fitting and other fields.Because of the easy acquisition of single photograph,three-dimensional face reconstruction based on single image has become a hot research topic.Face texture generated from the image is an important part of realizing realistic and aesthetic feeling in the process of face reconstruction.However,there are still some problems in face beautification in the current texture generation algorithms,such as skin color correction and eyebrow restoration.In this paper,the relevant research is on the above issues,and the main research contents include:(1)Face feature extraction.In this paper,the traditional Active Shape Model algorithm is improved and a multi-region ASM algorithm is proposed.Firstly,the cheek contour and the facial features are trained and modeled,then the deformation models are trained for different regions,and the feature points are extracted through the process from coarse to fine.(2)Color conversion and highlight removal.In order to correct the skin color of the human face,the method of editing and propagating combined with locally linear embedding algorithm is used to transform the skin color.The skin color region is extracted by skin color ellipse model,and the region that does not need to be converted is corrected.Then a bi-color reflection model is established to remove the specular reflection component of the face area and eliminate the high light of the face.(3)Seamless fusion and eyebrow repair.In order to reconstruct the occluded eyebrow contour,a method combining Li model and corner detection is proposed to reconstruct the semi-occluded eyebrow's contour.Finally,the facial regions of the beautified face image and the standard skin color image are fused by Poisson to generate the final face texture map.After obtain the texture map,through the texture mapping technology,a realistic and beautified three-dimensional face model is obtained.Experiments show that the proposed algorithm can be applied to face images taken under different complex background and illumination conditions.The skin color of the reconstructed three-dimensional face model is the same as that of the body model,and the occluded eyebrows are well reconstructed.The algorithm has a fast processing speed and can be applied to real-time interactive products.Finally,this paper summarizes and prospects the future work.