Efficient and synchronous three-dimensional facial animation generation method and system

Abstract

The invention relates to the technical field of image recognition and processing, and discloses an efficient and synchronous three-dimensional facial animation generation method, which comprises the following steps of S1, recognizing facial features from a high-resolution image through a deep learning model, and extracting geometric information and texture information; s2, generating a three-dimensional face model based on the geometric information, and constructing a dynamic expression parameterized model; s3, extracting pitch, speech speed and emotion characteristics from the input audio file, and generating expression driving data synchronized with the time axis; s4, combining the generated dynamic expression parameterized model with a high-resolution texture, and adjusting texture mapping in real time; and S5, rendering and outputting a three-dimensional facial animation synchronized with the audio. According to the invention, through algorithm optimization and hardware simplification, automation and intellectualization of a three-dimensional animation production process are realized, production cost is effectively reduced, and dependence on expensive equipment and professional animators is significantly reduced by means of optimization of a software algorithm and utilization of cloud computing resources.

Images