Method for modeling non-linear three-dimensional human face based on single sheet image

Abstract

The invention discloses a method for nonlinearly constructing the three-dimension model of a human face from a single image under the support from a three-dimension human face database. A group of characteristic points are extracted first according to an input image, and then, the parameters of head gestures in the input image are automatically estimated according to an average human face model in the human face base; after the gesture is restored, a neighborhood is determined in the three-dimension human face database for a given human face image according to a similarity measuring function; wherein, the value of the neighborhood is adaptively determined through a cross verification algorithm; an individualized three-dimension human face model is synthesized through a neighborhood interpolation algorithm; finally, a texture mapping is solved through the extracted characteristic points. In the method for constructing the three-dimension model of the human face, a high-precision three-dimension human face animation can be obtained and applied to separate single image under different experimental conditions so as to synthesize the three-dimensional model, thus ensuring a sense of reality of the human face animation; meanwhile, a verbose of manual division of the human face lacune area can be avoided; the production efficiency of the human face animation is enhanced; the method is more stable and efficient relative to the iterative algorithm in the calculating process of generating the three-dimension human face model; furthermore, the generated three-dimension human face model has textural information; a high-precision three-dimension human face animation with the sense of reality can be generated in real time.

Description

technical field [0001] The invention relates to the technical field of computer three-dimensional animation, in particular to a nonlinear three-dimensional face modeling method based on a single image. Background technique [0002] The earliest work on 3D facial animation technology began in 1972. Subsequent scholars have done a lot of work in order to generate realistic and vivid 3D facial animation. However, due to the complex anatomical structure of the human face, the subtle non-rigid movements are difficult to model mathematically, and at the same time, people's sensitivity to the appearance of the human face makes this topic very difficult. At present, some progress has been made in the work of realistic 3D facial animation, which can basically be divided into the following categories: [0003] Interpolation-based methods: The earliest facial animations were based on interpolation methods. First, the animator makes some specific facial expression keyframes, and then...

AI Technical Summary

Problems solved by technology

This approach allows for realistic and vivid animation of a specific character, but the resulting animation is difficult to transfer to other characters

Images